Brennan's conjecture: $L^s$-integrability of the derivative of a conformal map for $4/3<s<4$, equivalently boundedness of the induced composition operators on homogeneous Sobolev spaces

Brennan's conjecture (1978) asserts that for every conformal homeomorphism $\varphi:\Omega\to\mathbb{D}$ of a simply connected plane domain with nonempty boundary, $\iint_\Omega|\varphi'|^s\,d\mu<\infty$ for all $4/3<s<4$; equivalently, every univalent $f$ on the unit disk satisfies $\iint_\mathbb{D}|f'|^p\,dA<\infty$ for all $-2<p<2/3$, and equivalently the universal integral means spectrum satisfies $B(-2)=1$. Gol'dshtein and Ukhlov (2012) proved that the conjecture is equivalent to boundedness, for every $p\in(2,\infty)$, of the composition operators $\varphi^*:L^1_p(\mathbb{D})\to L^1_{q(p,s)}(\Omega)$, $q(p,s)=ps/(p+s-2)$, generated by conformal homeomorphisms. The conjecture remains open: the best unconditional results cover only about $s\le 3.752$ ($p\ge -1.752$; Hedenmalm-Shimorin 2005, marginally improved by Sola 2006), and 2024-2026 work still establishes it only for special classes (holomorphic semigroups, univalent rational functions, polygonal targets, polynomial basins of attraction).

Background

A conformal (univalent, i.e., holomorphic and injective) map between plane domains has a nowhere-vanishing complex derivative, and the Riemann mapping theorem guarantees, for every simply connected domain $\Omega\subset\mathbb{C}$ with nonempty boundary ($\Omega\ne\mathbb{C}$), a conformal homeomorphism $\varphi:\Omega\to\mathbb{D}$ onto the unit disk $\mathbb{D}$. Brennan's conjecture, raised by J. E. Brennan in 1978 in connection with the classical problem of mean approximation by analytic functions, asks how far the derivative of such a map is integrable.

The Koebe distortion theorem gives, for a univalent $f:\mathbb{D}\to\mathbb{C}$ normalized by $f(0)=0$ and $f'(0)=1$, the sharp pointwise bounds $(1-r)/(1+r)^3\le|f'(z)|\le(1+r)/(1-r)^3$ at $|z|=r$; both extremes are attained by rotations of the Koebe function $\kappa(z)=z/(1-z)^2$, which maps $\mathbb{D}$ onto the plane slit along $(-\infty,-1/4]$. Integrating the pointwise bounds shows $f'\in L^p(\mathbb{D},dA)$ for $-1<p<2/3$ (here $dA$ is planar Lebesgue measure), while $\iint_\mathbb{D}|\kappa'|^p\,dA<\infty$ holds exactly for $p<2/3$, so the positive endpoint $2/3$ cannot be improved. The entire difficulty of Brennan's conjecture lies on the negative side of the exponent range: negative moments $|f'|^p$ blow up where the derivative is small, i.e., near boundary points that correspond to long fjords or cusps of the image domain, and the question is how singular the boundary geometry of a simply connected domain can be. Brennan proved a range slightly beyond the distortion bound and conjectured the sharp statement, which in the convention of the source paper (integration over $\Omega$ for $\varphi:\Omega\to\mathbb{D}$, exponent $s=2-p$) reads $\iint_\Omega|\varphi'|^s\,d\mu<\infty$ for all $4/3<s<4$; the two conventions are equivalent by the change of variables $w=f(z)$. Both endpoints of the range are optimal: the slit plane $\mathbb{C}\setminus(-\infty,-1/4]$ forces $s>4/3$, and a slit disk shows the endpoint $s=4$ (equivalently $p=-2$) cannot be attained.

A second equivalent language is the integral means spectrum: for univalent $f$, let $\beta_f(t)$ be the infimum of $\beta$ with $\int_0^{2\pi}|f'(re^{i\theta})|^t\,d\theta=O((1-r)^{-\beta})$, and let $B(t)=\sup_f\beta_f(t)$ be the universal integral means spectrum. It is known that $B(t)=3t-1$ for $t\ge 2/5$, and Carleson and Makarov proved that $B(t)=|t|-1$ for all $t$ below some constant $\tau^*\le -2$. Brennan's conjecture is exactly the assertion $\tau^*=-2$, that is, $B(-2)=1$. In this form the known partial results read: the distortion bound gives only $p>-1$; refinements through Carleson-Makarov-type estimates and Bertilsson's thesis (1999) reached about $p\ge -1.421$; Shimorin (2005) proved $B(-1)\le 0.420$ and $B(-2)\le 1.246$; and the method of Hedenmalm and Shimorin (Duke Math. J. 2005), based on weighted Bergman spaces (spaces of holomorphic functions that are $L^p$-integrable with respect to area measure) and area-type estimates, marginally refined by Sola (2006), gives the best bounds to date, which translate back to integrability for $p\ge -1.752$, i.e., $s\le 3.752$ in the source convention. No improvement beyond roughly $p\gtrsim -1.75$ has been obtained since.

The source paper Gol'dshtein and Ukhlov (2012) adds a third, operator-theoretic formulation. The homogeneous Sobolev space $L^1_p(\Omega)$, $1\le p\le\infty$, is the space of locally integrable weakly differentiable functions with finite seminorm $(\iint_\Omega|\nabla f|^p\,d\mu)^{1/p}$; a homeomorphism $\varphi:\Omega\to\Omega'$ generates a composition operator $\varphi^*:L^1_p(\Omega')\to L^1_q(\Omega)$, $\varphi^*(f)=f\circ\varphi$, when this map is well defined and bounded. Their composition theorem states that for a conformal $\varphi:\Omega\to\Omega'$ this holds exactly when $\iint_\Omega|\varphi'|^{(p-2)q/(p-q)}\,d\mu<\infty$ ($1\le q<p<\infty$). Setting $(p-2)q/(p-q)=s$, i.e., $q=q(p,s)=ps/(p+s-2)$, yields their Equivalence Theorem: Brennan's conjecture holds for the exponent $s$ if and only if every conformal homeomorphism $\varphi:\Omega\to\mathbb{D}$ generates a bounded composition operator $\varphi^*:L^1_p(\mathbb{D})\to L^1_{q(p,s)}(\Omega)$ for every $p\in(2,\infty)$. For $p<2$ one gets $q(p,s)>p$, so no operators of this type exist, and the authors state that for $1<p<2$ they have no information; in a follow-up they propose the appropriate replacement, namely bounded composition operators in the inverse direction for $4/3<p<2$, which is again conjecturally equivalent to the same conjecture. The conjecture has been verified for special classes of maps and domains (starlike domains; continuous semigroups of holomorphic self-maps; univalent rational functions and maps onto bounded polygons; basins of attraction in complex dynamics), and the Gol'dshtein-Ukhlov equivalence underlies a program of sharp weighted Sobolev embeddings and spectral estimates for degenerate $p$-Laplace operators on planar domains.

Problem Statement

Prove or refute Brennan's conjecture: for every simply connected plane domain $\Omega\subset\mathbb{C}$ with nonempty boundary and every conformal homeomorphism $\varphi:\Omega\to\mathbb{D}$ onto the unit disk, $\iint_\Omega|\varphi'(z)|^s\,d\mu<\infty\quad\text{for all }4/3<s<4,$ where $\mu$ is planar Lebesgue measure. By the change of variables $p=2-s$ this is equivalent to: for every univalent function $f:\mathbb{D}\to\mathbb{C}$, $\iint_\mathbb{D}|f'(w)|^p\,dA(w)<\infty\quad\text{for all }-2<p<2/3,$ where $dA$ denotes planar Lebesgue measure, and to each of the following equivalent forms: the universal integral means spectrum satisfies $B(-2)=1$ (equivalently, the Carleson-Makarov constant $\tau^*$ equals $-2$); and, in the source paper's operator-theoretic form, every conformal homeomorphism $\varphi:\Omega\to\mathbb{D}$ generates a bounded composition operator $\varphi^*:L^1_p(\mathbb{D})\to L^1_{q(p,s)}(\Omega)$, $q(p,s)=ps/(p+s-2)$, for every $p\in(2,\infty)$ (in this clause $p$ denotes the Sobolev exponent of the composition-operator framework, not the area-integrability exponent of the preceding formulation) and every $s\in(4/3,4)$. Both endpoints of the exponent range are optimal (the slit plane forces $s>4/3$, and a slit disk shows $s=4$ cannot be attained), so the target is exactly the full interval and not a sub-range. The conjecture is refuted precisely by an explicitly specified simply connected domain with nonempty boundary together with an exponent in the stated range for which the corresponding integral diverges.

References

1. V. Gol'dshtein, A. Ukhlov, Brennan's conjecture for composition operators on Sobolev spaces, Eurasian Math. J. 3 (2012), No. 4, 35-43; arXiv:1212.3999, https://arxiv.org/abs/1212.3999 (Zbl 1281.30008)
2. J. E. Brennan, The integrability of the derivative in conformal mapping, J. London Math. Soc. (2) 18 (1978), 261-272, doi:10.1112/jlms/s2-18.2.261, https://doi.org/10.1112/jlms/s2-18.2.261
3. L. Carleson, N. G. Makarov, Some results connected with Brennan's conjecture, Arkiv for Matematik 32 (1994), No. 1, 33-62, doi:10.1007/BF02559522, https://doi.org/10.1007/BF02559522
4. H. Hedenmalm, S. Shimorin, Weighted Bergman spaces and the integral means spectrum of conformal mappings, Duke Math. J. 127 (2005), No. 2, 341-393, doi:10.1215/S0012-7094-04-12725-3, https://doi.org/10.1215/S0012-7094-04-12725-3 (arXiv:math/0406345)
5. S. Shimorin, Weighted composition operators associated with conformal mappings, in: Quadrature domains and their applications (The Harold S. Shapiro anniversary volume), Oper. Theory Adv. Appl. 156, Birkhauser, Basel, 2005, 217-237, doi:10.1007/3-7643-7316-4_11, https://doi.org/10.1007/3-7643-7316-4_11 (Zbl 1092.47028)
6. D. Bertilsson, On Brennan's conjecture in conformal mapping, Doctoral thesis, Royal Institute of Technology, Stockholm, 1999 (ISBN 91-7170-468-X; Zbl 0949.30013, https://zbmath.org/?q=an%3A0949.30013)
7. A. Sola, An estimate of the universal means spectrum of conformal mappings, Comput. Methods Funct. Theory 6 (2006), No. 2, 423-436, doi:10.1007/BF03321620, https://doi.org/10.1007/BF03321620
8. H. Hedenmalm, A. Sola, Spectral notions for conformal maps: a survey, Comput. Methods Funct. Theory 8 (2008), No. 2, 447-474, doi:10.1007/BF03321698, https://doi.org/10.1007/BF03321698
9. K. Barański, A. Volberg, A. Zdunik, Brennan's conjecture and the Mandelbrot set, Internat. Math. Res. Notices 1998 (1998), No. 12, 589-600, doi:10.1155/S1073792898000373, https://doi.org/10.1155/S1073792898000373
10. Ch. Pommerenke, Boundary behaviour of conformal maps, Grundlehren der mathematischen Wissenschaften 299, Springer-Verlag, Berlin, 1992, doi:10.1007/978-3-662-02770-7, https://doi.org/10.1007/978-3-662-02770-7
11. V. Gol'dshtein, A. Ukhlov, Sobolev homeomorphisms and Brennan's conjecture, Comput. Methods Funct. Theory 14 (2014), 247-256, doi:10.1007/s40315-014-0065-z, https://doi.org/10.1007/s40315-014-0065-z (arXiv:1309.1940)
12. V. Gol'dshtein, A. Ukhlov, Universal conformal weights on Sobolev spaces, arXiv:1302.4054, https://arxiv.org/abs/1302.4054
13. V. Gol'dshtein, V. Pchelintsev, A. Ukhlov, Spectral estimates of the p-Laplace Neumann operator and Brennan's conjecture, arXiv:1701.05143, https://arxiv.org/abs/1701.05143
14. A. Aleman, A. Kouroupis, Brennan's conjecture holds for semigroups of holomorphic functions, arXiv:2409.15074, https://arxiv.org/abs/2409.15074
15. J. Jin, Complex exponential integral means spectra of univalent functions and the Brennan conjecture, arXiv:2512.09330, https://arxiv.org/abs/2512.09330
16. Y. Zheng, Brennan conjecture for basin of attraction at infinity, arXiv:2604.12240, https://arxiv.org/abs/2604.12240
17. O. Tošić, Confirming Brennan's conjecture numerically on a counterexample to Thurston's $K=2$ conjecture, Experiment. Math. 33 (2024), No. 3, 422-436, doi:10.1080/10586458.2022.2153287, https://doi.org/10.1080/10586458.2022.2153287 (arXiv:2203.02577)
18. I. Prause, Quasidisks and twisting of the Riemann map, Int. Math. Res. Not. IMRN 2019 (2019), No. 7, 1936-1954, doi:10.1093/imrn/rnx173, https://doi.org/10.1093/imrn/rnx173 (arXiv:1611.06734)