Extremal quasiconformality vs bounded rational approximation

We show that, on most of the hyperbolic simply connected domains, the weighty bounded rational approximation in a natural sup norm is possible only for a very sparse set of holomorphic functions (in contrast to the integral approximation). The obstructions are caused by the features of extremal quas...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Український математичний вісник
Datum:2019
1. Verfasser: Krushkal, S.L.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Інститут прикладної математики і механіки НАН України 2019
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/169439
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Extremal quasiconformality vs bounded rational approximation / S.L. Krushkal // Український математичний вісник. — 2019. — Т. 16, № 2. — С. 181-199. — Бібліогр.: 26 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859650019530899456
author Krushkal, S.L.
author_facet Krushkal, S.L.
citation_txt Extremal quasiconformality vs bounded rational approximation / S.L. Krushkal // Український математичний вісник. — 2019. — Т. 16, № 2. — С. 181-199. — Бібліогр.: 26 назв. — англ.
collection DSpace DC
container_title Український математичний вісник
description We show that, on most of the hyperbolic simply connected domains, the weighty bounded rational approximation in a natural sup norm is possible only for a very sparse set of holomorphic functions (in contrast to the integral approximation). The obstructions are caused by the features of extremal quasiconformality. The paper is devoted to the 100th anniversary of Georgii Dmitrievich Suvorov, my first university adviser and teacher. He was an outstanding mathematician and a widely talented, extremely great human being.
first_indexed 2025-12-07T13:32:42Z
format Article
fulltext Український математичний вiсник Том 16 (2019), № 2, 181 – 199 Extremal quasiconformality vs rational approximation Samuel L. Krushkal (Presented by V. Gutlyanskĭı) The paper is devoted to the 100th anniversary of Georgii Dmitrievich Suvorov, my first university adviser and teacher. He was an outstanding mathematician and a widely talented, extremely great human being. Abstract. We show that on most of the hyperbolic simply connected domains the weighty bounded rational approximation in a natural sup norm is possible only for a very sparse set of holomorphic functions (in contrast to integral approximation). The obstructions are caused by the features of extremal quasiconformality. 2010 MSC. Primary: 30C62, 30C75, 30E10; Secondary: 30F45, 30F60, 32G15. Key words and phrases. Rational approximation, holomorphic func- tion, quasiconformal maps, quasicircles, universal Teichmüller space, Schwarzian derivative, Strebel point, Grunsky coefficients. 1. Results This paper gives a link of geometric function theory to weighty boun- ded rational approximation of holomorphic functions in sup norms and shows how the intrinsic features of extremal quasiconformal maps and universal Teichmüller space provide strong obstructions to such an ap- proximation. The situation is completely different from the integral ap- proximation. 1.1. Introductory remarks The classical directions in rational interpolation of holomorphic func- tions investigated by many authors concern mainly the uniform inter- polation of functions holomorphic in the inner points of the closed set Received 02.06.2019 ISSN 1810 – 3200. c⃝ Iнститут прикладної математики i механiки НАН України 182 Extremal quasiconformality vs bounded rational... X on the Riemann sphere Ĉ = C ∪ {∞} and continuous on X by ra- tional functions with poles off X and its combination with interpola- tion (see, e.g., [26]). The second approach was originated by Walsh (see [2,25]) and has recently been extended in [11] to the functional space A−∞ = ∪ q>0A −q over the Dini domains (with topology of the inductive limit), where A−q(D) is the Banach space of holomorphic functions in a domain D with norm ∥f∥ = supD δD(z)|f(z)|; here δD(z) = dist(z, ∂D) denotes the Euclidean distance from the point z ∈ D to the boundary. More generally, one considers a space F of holomorphic functions f in a domain D ⊂ Ĉ. For any given collections of points An = {anj}nj=0 ⊂ D, Bn = {bnj}nj=1 ⊂ Ĉ \D (n = 1, 2, . . . ), there exists a unique rational function rn,f of degree n, with poles at Bn, interpolating to f at An, counting multiplicities. The problem is to select these collections An and Bn so that for all f the interpolants rn,f converge to f in the topology of F . It is established in [11] (by sweeping out appropriate measures and application of the potential methods) that for f ∈ A−q(D), q > 0, the interpolants rn,f are convergent (under appropriate conditions) to f in A−q′ norm, where q′(q) ≫ q. 1.2. Weakened rational approximation in sup norm Let L be an oriented quasiconformal Jordan curve (quasicircle) on the Riemann sphere Ĉ = C ∪ {∞} with the interior and exterior domains D and D∗, and let p ≥ 2 be an integer. Denote by λD(z)|dz| the hyperbolic metric of D of Gaussian curvature −4 and consider the Banach spaces Ap(D), Bp(D) of holomorphic functions (quadratic differentials) φ with norms ∥φ∥Bp = sup D λD(z) −p|φ(z)|, ∥φ∥Ap = ∫∫ D λD(z) 2−p|φ(z)|dxdy, respectively; due to [5], the space Bp is dual to Ap. Note that for simply connected domains (with more than one bound- ary points) not containing inside the infinite point, 1 4 ≤ λD(z)δD(z) ≤ 1, (1) where the right hand inequality follows from the Schwarz lemma and the left from Koebe’s 1 4 theorem (so the spaces A−p mentioned above are obtained by renormalization of Bp). We also shall use the notations S. L. Krushkal 183 D = {z : |z| < 1}, D∗ = {z ∈ Ĉ : |z| > 1}; H = {z : Im z > 0}, H∗ = {z : Im z < 0}. We start with the following general Theorem 1. Let D ⊂ Ĉ be a domain with quasiconformal boundary L. Then for any function f ∈ Bp(D) there exists a sequence of rational functions with poles of order two on L of the form rn(z) = n∑ 1 cj (z − aj)2 , n∑ 1 |cj | > 0, (2) such that lim n→∞ ∥rn − φ∥Bp+1(D) = 0. In the case of the half-plane (or disk), this theorem is strengthened as follows. Theorem 2. For any φ ∈ Bp(H) there exists a sequence of rational functions (2) with real poles aj and real coefficients cj, convergent to φ in Bp+1(H). One can see from the proof of Theorem 1 that the weight exponent p + 1 is not sharp; though it is not clear whether this exponent can be replaces by p+ ϵ(p) with 0 < ϵ(p) < 1. The limit case ϵ = 0 has intrinsic interest. Then the assertion on convergence fails for ϵ = 0, because the space B2 of bounded holomorphic quadratic differentials φdz2 is not separable for any Riemann surface of infinite genus, and similar for all Bp. This was established by functional-analytic methods but can be es- tablished also from geometric features generated by Thurston’s theorem on existence of uncountable many conformally rigid domains (see [3,24]). Such domains correspond to the isolated points of the set U\T in B2(H), where U denotes the set of the Schwarzian derivatives Sw(z) = (w′′(z) w′(z) )′ − 1 2 (w′′(z) w′(z) )2 of all univalent functions on H and T is the universal Teichmüller space modeled by a bounded domain in B2(H) (formed by f having quasicon- formal extension to Ĉ). Note also that any φ ∈ B2(D) can be regarded as the Schwarzian of a locally univalent function in D and determines this function up to a Moebius transformation of Ĉ. 184 Extremal quasiconformality vs bounded rational... 1.3. Main theorems Our aim is to show that on most of the hyperbolic simply connected domains the rational approximation in B2 is possible only for a very sparse subset of functions. Theorem 3. For any simply connected domain D ⊂ Ĉ with quasiconfor- mal boundary L, whose conformal mapping function gD : H → D satisfies ∥SgD∥B2 < 1/2, the set of functions φ ∈ B2(D) approximated in B2 norm by general rational functions with poles of order two on L, rn(z) = n∑ 1 cj (z − aj)2 + n∑ 1 c′j z − aj , n∑ 1 |cj | > 0, (3) is nonwhere dense in the space B2(D). This theorem is a consequence of some deep results concerning the extremal maps and the Grunsky operator given by Theorem 4. There exists a constant c0 > 0 such that for any simply connected domain D ⊂ Ĉ with quasiconformal boundary L and such that ∥SgD∥B2 < 1/2, and for any rational function rn with poles of order two on L of the form (3) and norm ∥rn∥B2(D) < c0, we have the equalities κD(w) = k(w) = ∥rn∥B2(D), (4) where κD(w) and k(w) denote the Grunsky and Teichmüller norms of (appropriately normalized) univalent solution w : D → Ĉ of the Schwar- zian equation Sf = rn. Note that the indicated constant c0 does not depend on D. In the case of a disk (half-plane), one can take c0 = 1/2. Theorem 4 shows that all Schwarzians φ = rn with ∥rn∥B2 < c0 are not Strebel points in the universal Teichmüller space T (in other words, the conformal maps f with Sf = rn do not have Teichmüller extremal extensions onto the complementary domain D∗ = Ĉ \ D). Hence, such points cannot be dense in B2(D). Note also that, in view of the first equality in (4), the dilatation k(f) is attained on the squares of holomorphic abelian differentials ωdz on D. Theorem 3 is an immediate consequence of the equalities (4) in view of either of two basic results on openness and density in the universal S. L. Krushkal 185 Teichmüller space T: the result of [14] for points with non-equal Te- ichmüller and Grunsky norms and Lakic’s result [19] on the density of Strebel points (in arbitrary Teichmüller space). Indeed, it suffices to establish the assertion of Theorem 3 for φ ∈ B2(D) with ∥φ∥ < c0. Both Teichmüller and Grunsky norms are contin- uous on the universal Teichmüller space T modelled as a bounded domain in B2 containing the origin. Hence, the equality (4) is preserving also for the limit function f = lim rn of any sequence of rational functions in B2. But this equality implies that any f ∈ B2 with sufficiently small ∥f∥ generates the Beltrami coefficient µ(z) = 1 2 |z − z|2f ◦ gD(z), where gD is a conformal map of H onto domain D (called harmonic), and this coefficient is extremal in its equivalence class. On the other side, it is not of Teichmüller type. The latter is impossible in view of the indicated above openness re- sults. 1.4. Generalization of Theorems 3 and 4 One can see from the proof of Theorem 4 in Section 5 that actually its arguments are valid for an arbitrary meromorphic function φ on C having poles of order two which are located on a quasicircle L passing through the infinite point and accumulate to this point. This gives the following extension of the above theorems. Theorem 5. For any simply connected domain D ∈ Ĉ with quasicon- formal boundary L passing through the infinite point whose conformal mapping function satisfies ∥SgD∥B2 < 1/2, the subspace M2 in B2(D) formed by meromorphic functions φ on C with poles of order two, which are located on L and accumulate to ∞, is nonwhere dense in the space B2(D). If ∥φ∥B2(D) is sufficiently small, so that the Schwarzian equation Sw = φ has a univalent solution w(z) on D, then κD(w) = k(w) = ∥φ∥B2(D). 2. Proof of Theorem 1 First note that Ap(D) ⊂ BpD), and for any φ ∈ Ap(D), ∥φ∥Bp ≤ 4 π ∥φ∥Ap . (5) 186 Extremal quasiconformality vs bounded rational... Indeed, since both these norms are conformally invariant, it suffices to verify (5) for D = H. Then λD(z) = 1/(2y), and applying the mean inequality for holomorphic functions, one gets |f(z)| ≤ 1 πy2 ∫∫ |z−ζ|≤y |f(ζ)|dξdη ≤ (2y)2−p πy2 ∫∫ |z−ζ|≤y (2η)2−pdξdη (here ζ = ξ + iη, z ∈ D, η ≤ 2y), which yields (5). It follows also that for any φ ∈ Ap(D), lim ρ→0 sup δD(z)≥ρ λpD(z)|φ(z)| = 0. (6) Without loss of generality, one can assume that the boundary curve L ∋ ∞ and 0 ∈ D. For any φ ∈ Bp(D), |φ(z)| ≤ ∥φ∥BpλD(z) −p ≍ ∥φ∥BpδD(z) −p; hence it belongs to Ap+1 and its integral Iφ(z) = ∫ z 0 φ(ζ)dζ belongs to Ap. By the Bers approximation theorem [4], there exists a sequence of rational functions r̃j(z) with simple poles on L and no other singularities, such that1 lim n→∞ ∥r̃j − Iφ∥Ap = 0. (7) Further, due to [5], for every ψ ∈ Bp(D) the following reproducing for- mula is valid: ψ(z) = −2p− 1 π ∫∫ D (ζ − h(ζ)2p−2(∂h(ζ)/∂ζ))ψ(h(ζ)) (ζ − z)2p dξdη; (8) here ζ 7→ h(ζ) is a quasiconformal reflection with respect to the quasicircle L = ∂D (i.e., orientation reversing quasiconformal automorphism of Ĉ mapping D onto its complementary domain and leaving fixed all points of L) which is uniformly bilipschizian on C, i.e., for all points z1, z2 ∈ C the inequality c−1 0 |z1 − z2| ≤ |h(z1)− h(z2)| ≤ c0|z1 − z2| 1This theorem is proved in [4] for the integrable holomorphic functions f ∈ A2; the proof for the weighted spaces Ap can be done along the same lines, see [12]. S. L. Krushkal 187 holds with some constant c0 > 1. Moreover, the numerator νψ(ζ) = −2p− 1 π (ζ − h(ζ))2p−2∂h(ζ) ∂ζ is estimated uniformly by |νψ(ζ)| ≤ c(c0)∥ψ∥B2(D)λD(ζ) 2−p. Applying (8) to ψ = r̃j−Iφ and differentiating both sides in z, one obtains from (7) that rj = r̃′j are convergent to φ in Bp+1(D), completing the proof of the theorem. � 3. Proof of Theorem 2 We apply the following result on integral approximation in the unit disk given in [12] improving for the disk the Bers approximation theorem mentioned above and also related to the theory of extremal quasiconfor- mal maps. Proposition 1. Let p and m be two integers such that p ≥ 2 and m ≥ 1. Then for any ψ ∈ Ap(D), there exists a sequence of rational functions r̃j which have only simple poles on the unit circle S1 and satisfy the condition Im[ζmr̃j(ζ)] = 0 on S1 (outside of the poles of r̃j), such that lim j→∞ ∥r̃j − ψ∥Ap = 0. For even m and ζ = eiθ ∈ S1, we have Im[r̃j(ζ)dζ m] = (−1)m/2 Im[ζmr̃j(ζ)]dθ m; (9) thus the above proposition can be reformulated as follows. Proposition 2. For any function ψ ∈ Ap(D), there exists a sequence of rational functions r̃j which have only simple poles on the circle S1 and satisfy the condition Im[r̃j(ζ)dζ 2m] = 0 on S1, such that lim j→∞ ∥r̃j − ψ∥Ap = 0. Take m = p and, applying the fractional linear map σ(z) = (z − i)/(z+ i) of the upper half-plane onto the unit disk, approximate similar to Theorem 1 the integrated functions σ∗Iψ = (Iψ ◦ σ)(σ′)4m−2 188 Extremal quasiconformality vs bounded rational... by the corresponding rational functions σ∗r̃j = (r̃j ◦ σ)(σ′)4m−2 ∈ Ap+1(H) having real poles and coefficients in view of (9). Now applying the repro- ducing formula for the upper half-plane, ψ(z) = 2p− 1 π ∫∫ D (ζ − ζ)2p−2ψ(ζ) (ζ − z)2p dξdη, one straightforwardly obtains the conclusion of Theorem 2. 4. Backgrounds of Theorem 3 As was mentioned, Theorem 3 relies on some intrinsic features of ex- tremal quasiconformal maps and the Grunsky operator. For convenience, we briefly describe here these underlying results. 4.1. Extremal quasiconformality Let L be a quasicircle passing through the points 0, 1,∞ which is the common boundary of two domains D and D∗. Take the unit ball of Beltrami coefficients supported on D∗, Belt(D∗)1 = {µ ∈ L∞(C) : µ|D = 0 ∥µ∥∞ < 1} and consider the corresponding quasiconformal automorphisms wµ(z) of the sphere Ĉ satisfying on C the Beltrami equation ∂w = µ∂w preserv- ing the points 0, 1,∞ fixed. We call the quantity k(w) = ∥µw∥∞ the dilatation of the map w. Take the equivalence classes [µ] and [wµ] letting the coefficients µ1 and µ2 from Belt(D∗)1 be equivalent if the corresponding maps wµ1 and wµ2 coincide on L (and hence on D). These classes are in one-to-one correspondence with the Schwarzians Swµ on D which fill a bounded domain in the space B2(D) modelling the universal Teichmüller space T = T(D) with the base point D. The quotient map ϕT : Belt(D∗)1 → T, ϕT(µ) = Swµ is holomorphic (as the map from L∞(D∗) to B2(D)). Its intrinsic Te- ichmüller metric is defined by τT(ϕT(µ), ϕT(ν))= 1 2 inf { logK ( wµ∗◦ ( wν∗ )−1) : µ∗ ∈ ϕT(µ), ν∗ ∈ ϕT(ν) } , S. L. Krushkal 189 It is the integral form of the infinitesimal Finsler metric FT(ϕT(µ), ϕ ′ T(µ)ν) = inf{∥ν∗/(1− |µ|2)∥∞ : ϕ′T(µ)ν∗ = ϕ′T(µ)ν} on the tangent bundle T T of T, which is locally Lipschitzian. We call the Beltrami coefficient µ ∈ Belt(D∗)1 extremal (in its class) if ∥µ∥∞ = inf{∥ν∥∞ : ϕT(ν) = ϕT(µ)} and call µ infinitesimally extremal if ∥µ∥∞ = inf{∥ν∥∞ : ν ∈ L∞(D∗), ϕ′T(0)ν = ϕ′T(0)µ}. Any infinitesimally extremal Beltrami coefficient µ is globally extremal (and vice versa), and by the basic Hamilton–Krushkal–Reich–Strebel the- orem the extremality of µ is equivalent to the equality ∥µ∥∞ = inf{| < µ,ψ >D∗ | : ψ ∈ A2(D ∗) : ∥ψ∥ = 1} (where A2(D ∗) is the subspace of L1(D ∗) formed by holomorphic func- tions on D∗) and the pairing ⟨µ, ψ⟩D∗ = ∫∫ D∗ µ(z)ψ(z)dxdy, µ ∈ L∞(D∗), ψ ∈ L1(D ∗) (z = x+iy). Let w0 := wµ0 be an extremal representative of its class [w0] with dilatation k(w0) = ∥µ0∥∞ = inf{k(wµ) : wµ|L = w0|L}, and assume that there exists in this class a quasiconformal map w1 whose Beltrami coefficient µA1 satisfies the inequality ess supAr |µw1(z)| < k(w0) in some ring domain R = D∗ \G complement to a domain G ⊃ D∗. Any such w1 is called the frame map for the class [w0], and the corresponding point in the universal Teichmüller space T is called the Strebel point. These points have the following important properties. Proposition 3. (i) If a class [f ] has a frame map, then the extremal map f0 in this class (minimizing the dilatation ∥µ∥∞) is unique and either a conformal or a Teichmüller map with Beltrami coefficient µ0 = k|ψ0|/ψ0 on D∗, defined by an integrable holomorphic quadratic differential ψ0 on D∗ and a constant k ∈ (0, 1) [23]. (ii) The set of Strebel points is open and dense in T [8, 19]. 190 Extremal quasiconformality vs bounded rational... The first assertion holds, for example, for asymptotically conformal (hence for all smooth) curves L. Similar results hold also for arbitrary Riemann surfaces (cf. [7, 8]). The boundary dilatation H(f) admits also a local version Hp(f) in- volving the Beltrami coefficients supported in the neighborhoods of a boundary point p ∈ ∂D. Moreover (see, e.g., [8, Ch. 17]), H(f) = supp∈∂DHp(f), and the points with Hp(f) = H(f) are called substan- tial for f and for its equivalence class. 4.2. The Grunsky–Milin inequalities Let D∗ ∋ ∞ be a simply connected domain with quasiconformal boundary and Σ0(D∗) denote the class of univalent Ĉ-holomorphic func- tions in D∗ with expansions f(z) = z + b0 + b1z −1 + . . . near z = ∞ admitting quasiconformal extensions to Ĉ. Their Grunsky–Milin coeffi- cients αmn are defined from the expansion − log f(z)− f(ζ) z − ζ = ∞∑ m,n=1 αmn χ(z)m χ(ζ)n , (10) choosing the branch of the logarithmic function which vanishes as z = ζ → ∞. Here χ denotes a conformal map of D∗ onto the disk D∗ so that χ(∞) = ∞, χ′(∞) > 0. Each coefficient αmn(f) in (10) is a polynomial of a finite number of the initial coefficients b1, b2, . . . , bm+n−1 of f ; hence it depends holo- morphically on Beltrami coefficients of extensions of f as well as on the Schwarzian derivatives Sf ∈ B2(D ∗). A theorem of Milin extending the Grunsky univalence criterion for the disk D∗ states that a holomorphic function f(z) = z+const+O(z−1) in a neighborhood of z = ∞ can be continued to a univalent function in the whole domain D∗ if and only if the coefficients αmn satisfy the inequality ∣∣∣ ∞∑ m,n=1 √ mn αmnxmxn ∣∣∣ ≤ 1 for any point x = (xn) from the unit sphere S(l2) of the Hilbert space of sequences x = (xn) with ∥x∥2 = ∞∑ 1 |xn|2 (cf. [10, 20, 22]). We call the quantity κD∗(f) = sup {∣∣∣ ∞∑ m,n=1 √ mn αmn xmxn ∣∣∣ : x = (xn) ∈ S(l2) } S. L. Krushkal 191 the Grunsky norm of f . The inequality κD∗(f) ≤ 1 is necessary and sufficient for univalence of f in D∗ (see [10,20,22]). In the canonical case D∗ = D∗, we have the classical Grunsky coefficients. Consider the set A2 2(D) = {ψ ∈ A2(D) : ψ = ω2} consisting of the integrable holomorphic functions on D having only zeros of even order and put αD(f) = sup {|⟨µ0, ψ⟩D| : ψ ∈ A2 2, ∥ψ∥A2(D) = 1}. The following proposition from [16] completely describes the relation be- tween the Grunsky and Teichmüller norms (more special results were obtained in [13,18]). Proposition 4. For all f ∈ Σ0(D∗), κD∗(f) ≤ k k + αD(f) 1 + αD(f)k , k = k(f), and κD∗(f) < k unless αD(f) = ∥µ0∥∞, (11) where µ0 is an extremal Beltrami coefficient in the equivalence class [f ]. The last equality is equivalent to κD∗(f) = k(f). If κD∗(f) = k(f) and the class of [f ] is a Strebel point, then µ0 is necessarily of the form µ0 = ∥µ0∥∞|ψ0|/ψ0 with ψ0 ∈ A2 2(D). Note that geometrically (11) means the equality of the Carathéodory and Teichmüller distances on the geodesic disk {ϕT(tµ0/∥µ0∥) : t ∈ D} in the universal Teichmüller space T. 5. Proof of Theorem 4 We first prove this theorem for D = H (and hence for the disk). In this canonical case, one gets a somewhat stronger result; moreover, the arguments are simpler and illustrate all underlying features. Now the poles of rn are real, and λH(z) = 1/|z − z| = 1/(2y). The assertion of the theorem follows from the next two lemmas. The first lemma ensures the existence for any rn ∈ B2(D) of a sequence of points zn ∈ D convergent to a boundary point a0 on which the supremum 192 Extremal quasiconformality vs bounded rational... of λ−2 D |rn| is attained (such a0 can be distinct from the poles of rn). The second lemma yields that this a0 must be an essential point for rn, and therefore this function represents a non-Strebel point. Of course, all this is valid to much more general functions from B2(D). A special case (the convex hull of fractions 1/(z − a)2 with real a) was considered in [15]. Lemma 1. For any simply connected domain D ⊂ Ĉ with quasiconformal boundary L and any rational function rn with poles of order two on L of the form (3), ∥rn∥B2(D) = lim sup z→L λD(z) −2|rn(z)|. (12) So, there is a boundary point z0 at which the maximal value of λD(z) −2|rn(z)| on D is attained. Proof. Consider first the case D = D, and let rn ∈ B2(D) satisfy lim sup |z|→1 (1− |z|2)2|rn(z)| < sup z∈D (1− |z|2)2|rn(z)|, (13) i.e., the polyanalytic function F (z) = (1− zz)2|rn(z)| with F (0) = rn(0) attains its maximal value on D at some inner point z0 ∈ D. Applying, if needed, the conformal automorphism z 7→ (z − z0)/(1− z0z) of D, one reduces the proof to the case z0 = 0. If r′n(0) = a ̸= 0, then rn(z) = rn(0)+az+. . . , and hence, for z = ρeiθ and small ρ > 0, max θ |rn(ρeiθ)| = |rn(0)|+ |a|ρ+O(ρ2). This yields max θ |F (ρeiθ)| = |F (0)|+ |a|ρ+O(ρ2) > |F (0)|, ρ→ 0, which contradicts the maximality of |F (z)| at z = 0. So, for such rational functions rn, the inequality (13) can never occur, and lim sup |z|→1 (1− |z|2)2|rn(z)| = sup z∈D (1− |z|2)2|rn(z)| = ∥rn∥B2(D). S. L. Krushkal 193 If r′n(0) = 0, we approximate this function by rational rn,ε with the same poles ak ∈ ∂D, replacing one of the coefficients ck by ck + ε so that r′n,ε(0) ̸= 0. Since at the point z0, where the function Fε(z) = (1− zz)2|rn,ε(z)| attains its maximal value, this value is positive, one can define in a neigh- borhood of z0 a single valued branch gn,ε(z) = √ rn,ε(z), and in this neighborhood Fε(z) = (1− zz)2gn,ε(z)gn,ε(z). Noting that both partial derivatives ∂zFε(z), ∂zFε(z) vanish at z0 and ∂zFε(z) = −2(1− zz)zgn,ε(z)gn,ε(z) + (1− zz)2gn,ε(z)g ′ n,ε(z), one obtains −2z0gn,ε(z0) + (1− z0z0)g ′ n,ε(z0) = 0, and therefore, g′n,ε(z0) = − 2z0 1− z0z0 gn,ε(z) ̸= 0. This yields, in the same manner as above, that every such rn,ε satisfies the equality (12). Since this equality remains valid in the limit as ε→ 0, the assertion of Lemma for D = D is established. The case of the generic quasidiskD is reduced to the above one, taking a conformal map χD function of D onto the disk D with χD(z0) = 0 and applying the above arguments to functions rn ◦ χ, which have the same properties as rn. This completes the proof of the lemma. � Note that rn(z) = O(1/z2) as z → ∞, so the quadratic differential rn(z)dz 2 has at the infinite point a pole of the second order. If the boundary of domain D contains z = ∞, then the maximal value in (12) can be obtained at this point (and accordingly, (1 − |ζ|2)2|rn(ζ)χD(ζ)| can attain its maximum at ζ = χD(∞)). Lemma 2. Let D be a simply connected domain on Ĉ with quasiconfor- mal boundary L and such that ∥SgD∥B2 < 1/2. There exists a constant c0 > 0 such that for any rational function rn with poles of order two on L of the form (3) and with ∥rn∥B2(D) < c0 194 Extremal quasiconformality vs bounded rational... the boundary points of D at which the maximal value in (12) is attained are substantial for extremal quasiconformal extensions of conformal im- mersions f : D → Ĉ generated by the Schwarzian equation Sf = rn on D. Proof. It is sufficient to prove the lemma for domains with boundaries containing ∞. We first consider the canonical case D = H for which a somewhat stronger result will be obtained. The equation Sf (z) = φ(z) defines the conformal immersion fφ : H → Ĉ determined uniquely by the requirement to preserve the points 0, 1,∞. By the Ahlfors–Weill theorem [1], every φ ∈ B2(H) with ∥φ∥ < 1/2 is the Schwarzian derivative Sf of a univalent function f in H, and f has a quasiconformal extension onto the lower half-plane H∗ with Betrami coefficient of the form µφ(z) = −2y2φ(z), φ = Sf (z = x+ iy ∈ H) (14) called the harmonic Beltrami coefficient (in the spirit of the Kodaira– Spencer deformation theory). Our aim is to show that for every rn with real poles aj of order two and ∥rn∥B2(H) < 1/2 the corresponding harmonic Beltrami coefficient µrn in H is extremal in its class, and κ(frn ◦ σ) = k(frn) = ∥rn∥B2(H∗), (15) where σ is the appropriate Moebius map of D∗ onto H∗. It suffices to establish the relations (15) for rn with sufficiently small norm. Pick the point a0 ∈ R at which the equality in (12) is attained, and two points x′, x′′ located on R in the left to all poles aj (so, µrn(z) = 0 on [x′, x′′]). We now establish that sup ∥ψ∥A2(H)=1 |⟨µrn , ψ⟩H| = sup ∥ψ∥ A2 2(H) =1 |⟨µrn , ψ⟩H| = Ha0(f), (16) which implies the equalities (15) and extremality of µrn in its class. Using the conformal map z = g(ζ) of the half-strip Π+ = {ζ = ξ + iη : ξ > 0, 0 < η < 1} onto H with g(x′) = 0, g(x′′) = 1, g(a0) = ∞, we pull-back µrn/b(a0) (where b(a0) is the local boundary dilatation at the point a0) to the Beltrami coefficient µ∗(ζ) := 1 Ha0 g∗(µrn)(ζ) = 1 Ha0 (µrn ◦ g)(ζ) g′(ζ)/g′(ζ) S. L. Krushkal 195 on Π+, which satisfies lim ξ→∞ |µ∗(ξ + iη)| = ∥µ∗∥∞ = 1 and has the limit function µ∗(ζ0) = lim ζ→ζ0∈∂Π+ µ∗(ζ) with µ∗(iη) = 0. (17) We claim that the sequence ωm(ζ) = 1 m e−ζ/m, m = 1, 2, . . . (ζ ∈ Π+), is degenerating for ν∗. First of all, these ωm belong to A2 2(Π+); ωm(ζ) → 0 uniformly on Π+ ∩ {|ζ| < M} for any M < ∞, and ∥ωm∥A2(Π+) = 1. Further, ⟨µ∗, ωm⟩Π+ = 1 m ∫∫ Π+ µ∗(ζ)ωm(ζ)dξdη = 1∫ 0 e−iη/mdη ( 1 m ∞∫ 0 µ∗(ξ + iη)e−ξ/mdξ ) . (18) The inner integral can be evaluated using the Laplace transform of µ∗ in ξ. Integrating by parts and applying (17), one obtains ∞∫ 0 ∂µ∗(ξ + iη) ∂ξ e−ξ/mdξ = 1 m ∞∫ 0 µ∗(ξ + iη)e−ξ/mdξ. On the other hand, Abel’s theorem for the Laplace transform yields that the nontangential limit lim s→0 ∞∫ 0 ∂µ∗(ξ + iη) ∂ξ e−sξdξ = ∞∫ 0 ∂µ∗(ξ + iη) ∂ξ dξ = µ∗(∞)− µ∗(iη); hence, lim m→∞ 1 m ∞∫ 0 µ∗(ξ + iη)e−ξ/mdξ = µ∗(∞). By Lebesgue’s theorem on dominated convergence, the iterated integral in (18) is estimated as follows lim m→∞ |⟨ν∗, ωm⟩Π+ | = ∣∣∣ 1∫ 0 dη lim m→∞ 1 m ∞∫ 0 µ∗(ξ + iη)e−ξ/mdξ ∣∣∣ = 1. (19) 196 Extremal quasiconformality vs bounded rational... Since by (12), the left-hand side equals to ∥µ∗∥∞ and all functions ωm belong to A2 2(Π+), this proves our claim. Now, applying the inverse conformal map ζ = g−1(z) : Π+ → H, one obtains the degenerating sequence {ψm = (ωm ◦ g−1)(g′)−2} ⊂ A2 2(H), for the initial Beltrami coefficient µrm on H. By (19), lim m→∞ |⟨µrm , ψm⟩H| = ∥µrm∥∞ = 1, which implies, together with Lemma 1, the assertion of Theorem 4 for the half-plane. Let now D be the generic simply connected domain bounded by qua- sicircle L passing through 0, 1,∞ and such that the Schwarzian of the conformal map gD of H∗ onto D preserving the points 0, 1,∞ satisfies ∥SgD∥B2(H∗) < 1/2. Given a rational function rn(z) of the form (3) with poles on L, we consider the univalent solution w = fn(z) of the equation Sw(z) = trn(z), z ∈ D and its composition with gD, taking t > 0 so small that Sfn◦gD = (Sf ◦ gD)(g′D)2 + SgD (20) also satisfies ∥Sfn◦gD∥B2(H∗) < 1/2. (21) The Beltrami coefficients of arbitrary quasiconformal extensions ĝD and f̂n of gD and fn, respectively, across the boundaries of their domains to Ĉ are related by µ f̂n ◦ ĝD = µ f̂n◦ĝ−1 D ◦ ĝD = µ f̂n◦ĝD − µĝD 1− µĝDµf̂n◦ĝD ∂ζ ĝD ∂ζ ĝD . In particular, using their Ahlfors–Weill extensions (14), one gets, in view of (20), µ f̂n ◦ ĝD = −2tη2rn ◦ gD(η) ∂ζ ĝD ∂ζ ĝD +O(t2) = −2tλ−2 D rn +O(t2), S. L. Krushkal 197 or equivalently, for z = gD(ζ), µ f̂ (z) = −2tλ−2 D (z)rn(z) +O(t2) as t→ 0. The remainders in the last two equalities are uniformly bounded in L∞ norm for all t for which the bound (21) is valid. We establish now that the harmonic Beltrami coefficient νtrn(z) = tλ−2 D (z)rn(z) (22) is infinitesimally extremal in its equivalence class. Indeed, taking again the point a0 ∈ L on which the upper limit (12) for chosen rn is attained and mapping the domain D conformally onto the half-strip Π+ so that a0 is going to ∞, one can repeat for fn the above arguments and derive from (16) and (19) that the boundary dilatation at a0 is equal to ∥rn∥B2(D). This yields that the Beltrami coefficient (22) is infinitesimally extremal in Belt(D)1 and, moreover, its norm is attained on functions from A2 2(D). As was mentioned in Section 4.1, such a Beltrami coefficient must be simultaneously globally extremal in its equivalence class. This implies the assertions of Lemma 2 and of Theorem 4, completing their proofs. � References [1] L. V. Ahlfors, G. Weill, A uniqueness theorem for Beltrami equations // Proc. Amer. Math. Soc., 13 (1962), 975–978. [2] A. Amboladze, H. Wallin, Rational interpolants with prescribed poles, theory and practice // Complex Var. Theory Appl., 34(4) (1997), 399–413. [3] K. Astala, Selfsimilar zippers, Holomorphic Functions and Moduli, vol. I (D. Drasin et al., eds.), Springer-Verlag, New York, 1988, pp. 61–73. [4] L. Bers, An approximation theorem // J. Anal. Math. 14 (1965), 1–4. [5] L. Bers, A non-standard integral equation with applications to quasiconformal mappings // Acta Math., 116 (1966), 113–134. [6] C. J. Earle, I. Kra, S. L. Krushkal, Holomorphic motions and Teichmüller spaces // Trans. Amer. Math. Soc., 944 (1994), 927–948. [7] C. J. Earle, Zong Li, Isometrically embedded polydisks in infinite dimensional Teichmüller spaces // J. Geom. Anal., 9 (1999), 51–71. [8] F. P. Gardiner, N. Lakic, Quasiconformal Teichmüller Theory // Amer. Math. Soc., 2000. [9] G. M. Golusin, Geometric theory of Functions of a Complex Variable, Transl. of mathematical monographs, vol. 26, Transl. of Geometricheskaya teoriya funk- tcii kompleksnogo peremennogo, 2nd ed, Amer. Math. Soc., Providence, RI, 1969. 198 Extremal quasiconformality vs bounded rational... [10] H. Grunsky, Koeffizientenbedingungen für schlicht abbildende meromorphe Funktionen // Math. Z., 45 (1939), 29–61. [11] A. Gustafsson, Approximation with rational interpolants in A−∞(D) // Com- put. Methods Func. Theory, DOI 10.1007/s40315-016-0187-6. [12] S. L. Krushkal, Quasiconformal Mappings and Riemann Surfaces, Wiley, New York, 1979. [13] S. L. Krushkal, Grunsky coefficient inequalities, Carathéodory metric and ex- tremal quasiconformsal mappings // Comment. Math. Helv., 64 (1989), 650– 660. [14] S. L. Krushkal, Strengthened Moser’s conjecture, geometry of Grunsky inequal- ities and Fredholm eigenvalues // Central European J. Math., 5(3) (2007), 551–580. [15] S. L. Krushkal, Rational approximation of holomorphic functions and geometry of Grunsky inequalities // Contemporary Mathematics, 455 (2008), 219–236. [16] S. L. Krushkal, Strengthened Grunsky and Milin inequalities // Contemp. Math., 667 (2016), 159–179. [17] R. Kühnau, Verzerrungssätze und Koeffizientenbedingungen vom Grun- skyschen Typ für quasikonforme Abbildungen // Math. Nachr., 48 (1971), 77– 105. [18] R. Kühnau, Wann sind die Grunskyschen Koeffizientenbedingungen hinre- ichend für Q-quasikonforme Fortsetzbarkeit? // Comment. Math. Helv., 61 (1986), 290–307. [19] N. Lakic, Strebel Points // Lipa’s Legacy, Contemporary Mathematics, 211, Amer. Math. Soc., Providence, RI, 2001, 417–431. [20] I. M. Milin, Univalent Functions and Orthonormal Systems, Transl. of Math- ematical Monographs, vol. 49, Transl. of Odnolistnye funktcii i normirovannie systemy, Amer. Math. Soc., Providence, RI, 1977. [21] Z. Nehari, The Schwarzian derivative and schlicht functions // Bull. Amer. Math. Soc., 55 (1949), 545–551. [22] Chr. Pommerenke, Univalent Functions, Vandenhoeck & Ruprecht, Göttingen, 1975. [23] K. Strebel, On the existence of extremal Teichmueller mappings // J. Anal. Math., 30 (1976), 464–480. [24] W. P. Thurston, Zippers and univalent functions, The Bieberbach Conjecture: Proceedings of the Symposium on the Occasion of its Proof (A. Baernstein II et al., eds.), Amer. Math. Soc., Providence, R.I., 1986, 185–197. [25] J. L. Walsh, Interpolation and Approximation by Rational Functions in the Complex Domain, 4th edn, Colloq. Publ., vol. XX, Amer. Math. Soc., Provi- dence, RI, 1965. S. L. Krushkal 199 [26] L. Zalcman, Analytic Capacity and Rational Approximation, Lecture Notes in Math., 50, Springer, Berlin, 1968. Contact information Samuel L. Krushkal Department of Mathematics, Bar-Ilan Uni- versity, Israel, Department of Mathematics, University of Virginia, Charlottesville, USA E-Mail: krushkal@math.biu.ac.il
id nasplib_isofts_kiev_ua-123456789-169439
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1810-3200
language English
last_indexed 2025-12-07T13:32:42Z
publishDate 2019
publisher Інститут прикладної математики і механіки НАН України
record_format dspace
spelling Krushkal, S.L.
2020-06-13T10:16:25Z
2020-06-13T10:16:25Z
2019
Extremal quasiconformality vs bounded rational approximation / S.L. Krushkal // Український математичний вісник. — 2019. — Т. 16, № 2. — С. 181-199. — Бібліогр.: 26 назв. — англ.
1810-3200
2000 MSC. Primary: 30C62, 30C75, 30E10; Secondary: 30F45, 30F60, 32G15
https://nasplib.isofts.kiev.ua/handle/123456789/169439
We show that, on most of the hyperbolic simply connected domains, the weighty bounded rational approximation in a natural sup norm is possible only for a very sparse set of holomorphic functions (in contrast to the integral approximation). The obstructions are caused by the features of extremal quasiconformality.
The paper is devoted to the 100th anniversary of Georgii Dmitrievich Suvorov, my first university adviser and teacher. He was an outstanding mathematician and a widely talented, extremely great human being.
en
Інститут прикладної математики і механіки НАН України
Український математичний вісник
Extremal quasiconformality vs bounded rational approximation
Article
published earlier
spellingShingle Extremal quasiconformality vs bounded rational approximation
Krushkal, S.L.
title Extremal quasiconformality vs bounded rational approximation
title_full Extremal quasiconformality vs bounded rational approximation
title_fullStr Extremal quasiconformality vs bounded rational approximation
title_full_unstemmed Extremal quasiconformality vs bounded rational approximation
title_short Extremal quasiconformality vs bounded rational approximation
title_sort extremal quasiconformality vs bounded rational approximation
url https://nasplib.isofts.kiev.ua/handle/123456789/169439
work_keys_str_mv AT krushkalsl extremalquasiconformalityvsboundedrationalapproximation