Milnor's problem: an effective criterion, in terms of $f$, for the parabolic or hyperbolic type of the complete graph $z=f(x,y)$

A complete simply connected Riemannian surface is of parabolic type if it admits no Green's function (equivalently, its Brownian motion is recurrent, or every harmonic function on it that is bounded above or below is constant) and of hyperbolic type otherwise; deciding which is the classical type problem. Milnor (1977) proved sharp Gaussian-curvature criteria deciding the type of a surface of revolution and posed two questions: whether his tests survive without rotational symmetry, and whether the type of a surface embedded in Euclidean 3-space as a graph $z=f(x,y)$ can be decided effectively. The first question is settled: Doyle (1988) and Grigor'yan (1987, 1999) extended the tests to surfaces without rotational symmetry, the parabolic test to all complete surfaces and the hyperbolic test to surfaces with a pole. The second question remains open and is the subject of this record: find a necessary and sufficient, effectively checkable criterion in terms of the defining function $f$ that decides the conformal type of the complete graph $z=f(x,y)$ over $\mathbb{R}^2$. Only partial sufficient conditions are known, chiefly the capacity-based integral tests of Keselman, Igonina, Kozlova and Paraskevopulo (2024) in the tradition of Miklyukov (1996), and between the known sufficient parabolic and sufficient hyperbolic regimes an undecided middle remains for general $f$.

Background

Every complete simply connected Riemannian surface $M$ — a two-dimensional manifold with a length metric, equivalently (by the uniformization theorem, which classifies simply connected Riemann surfaces) an open simply connected Riemann surface — is conformally equivalent to exactly one of the complex plane and the unit disk. Which of the two occurs is the conformal type of $M$, and it has equivalent analytic characterizations: $M$ is called parabolic if it admits no Green's function, i.e. no positive solution of $\Delta u=-\delta_p$ on $M\setminus\{p\}$ with a single logarithmic pole $p$ (here $\Delta$ is the Laplace operator of the surface), and hyperbolic otherwise; Brownian motion, the diffusion generated by half the Laplacian, is recurrent — visits every nonempty open set infinitely often — exactly in the parabolic case; and in capacity terms, where the capacity of a compact set $K\subset M$ is $\inf\{\int_M|\nabla h|^2\,d\sigma:\ h$ smooth, compactly supported, $h\ge 1$ in a neighborhood of $K\}$, $M$ is parabolic iff every compact subset has capacity zero. Deciding the type of a surface given by geometric data — the type problem, posed explicitly by Ahlfors (1935) and classical since the early twentieth century — is in general hard, because the type is a global conformal invariant not determined by local quantities.

For a surface of revolution, written in geodesic polar coordinates (coordinates $(r,\theta)$ where $r$ is geodesic distance from a fixed point and $\theta$ the direction of the geodesic) as $dr^2+g(r)^2d\theta^2$ with $g(r)>0$, the Gaussian curvature (the intrinsic curvature of the surface) is $K=-g''(r)/g(r)$, and Milnor (1977) gave an effective answer: if $K(r)\ge -1/(r^2\log r)$ for all sufficiently large $r$, the surface is parabolic; if $K(r)\le -(1+\varepsilon)/(r^2\log r)$ for some $\varepsilon>0$ and all large $r$, and $g(r)$ is unbounded, it is hyperbolic. The borderline scale $r^2\log r$ is sharp. Both proofs exploit rotational symmetry: the Laplacian separates in $(r,\theta)$, and parabolicity reduces to divergence of the one-dimensional integral $\int^\infty dr/g(r)$, the same integral that decides transience of a spherically symmetric resistor network. Milnor noted that a length criterion going back to Ahlfors (1935) and Nevanlinna (1940) — a complete surface is parabolic if $\int^\infty dr/L(r)=\infty$, where $L(r)$ is the length of the geodesic circle of geodesic radius $r$ (Grigor'yan 1999, Theorem 7.5) — already gives the parabolic test without symmetry, but that no analogous symmetry-free generalization of the hyperbolic test was known; and he asked for an effective way to decide the type of a surface embedded in Euclidean 3-space as a graph.

The symmetry limitation was removed by later work. Doyle (1988), in a paper deliberately carrying the same title as Milnor's, rederived both tests for simply connected surfaces without rotational symmetry from an electrical-network heuristic that treats the surface as a resistor network whose total resistance to infinity is finite exactly in the hyperbolic case; Grigor'yan (1987) made the extension rigorous in all dimensions for manifolds with a pole — a point $o$ whose cut locus (the set where geodesics leaving $o$ stop minimizing distance) is empty, so that global polar coordinates centered at $o$ exist. The final form (Grigor'yan 1999, Theorem 15.6) states: the parabolic test holds on any two-dimensional geodesically complete manifold, with no symmetry assumption at all, while the hyperbolic test holds for manifolds with a pole under the additional hypothesis $\limsup_{r\to\infty}L(r)>0$, which excludes cusp-like ends that are parabolic for trivial reasons; the survey credits both tests as first proved by Milnor for model manifolds (spherically symmetric ones) and then by Doyle for manifolds with a pole. A later refinement in this lineage — whether the Doyle–Grigor'yan non-parabolicity criterion is also necessary — was answered negatively in dimensions $d\ge 4$ by Bessa, Gimeno i Garcia, Pessoa and Setti (2024); dimensions 2 and 3 remain open, but this necessity question is not part of Milnor's original problem.

The graph question survives. An entire graph is the surface $M_f=\{(x,y,f(x,y)):(x,y)\in\mathbb{R}^2\}$ with the metric induced from Euclidean $\mathbb{R}^3$,

$ds^2=(1+f_x^2)\,dx^2+2f_xf_y\,dx\,dy+(1+f_y^2)\,dy^2,$

a complete surface homeomorphic to the plane. The vertical projection identifies $M_f$ with $\mathbb{R}^2$ as a smooth manifold but not conformally: the displayed metric is in general not a scalar multiple of the Euclidean metric, and its conformal class — the only thing the type depends on — is only indirectly controlled by $f$. Both outcomes occur, and neither trivially. If $f$ is Lipschitz, the projection is bilipschitz (it changes distances by at most a fixed factor), which preserves parabolicity, so $M_f$ is parabolic; if $f$ depends only on the polar radius $\rho=\sqrt{x^2+y^2}$, the parallel circles keep their Euclidean lengths and the Ahlfors–Nevanlinna criterion forces parabolicity no matter how fast $f$ grows (Keselman et al. 2024, Corollary 1). Hyperbolic entire graphs nevertheless exist: Huber (1959), Jenkins (1961) and Osserman (1961) constructed them, answering a question attributed to C. Loewner, and Keselman et al. (2024) note that each of those constructions verifies hyperbolicity by an argument specific to the construction rather than by a general condition. All known constructions let the gradient of $f$ become unbounded and oscillate strongly in the angular direction: for a graph, hyperbolicity can only be produced by the angular derivative $f_\varphi'$ outgrowing the radial scale, which is exactly what the sufficient conditions of Keselman et al. (2024) formalize. No growth test can decide the question, since type is orthogonal to size: parabolic surfaces can have arbitrarily fast volume growth of geodesic balls (Keselman et al. 2024, in the tradition of Miklyukov's capacity criteria).

The state of the art for the graph problem is a collection of one-sided sufficient conditions. The general framework is the capacity theory of Miklyukov (1996), which gives criteria for parabolicity and hyperbolicity in terms of capacities of boundary sets. Keselman, Igonina, Kozlova and Paraskevopulo (2024) — a paper that opens by recalling that Milnor posed the problem of finding effective conditions of parabolic and hyperbolic type for non-parametric surfaces — derive from upper and lower capacity estimates a sufficient condition for parabolicity (divergence of a radial integral whose integrand is an angular average built from the polar-coordinate derivatives $f_\rho'$ and $f_\varphi'$ of $f$) and a sufficient condition for hyperbolicity (a complementary integrability condition), and they construct a hyperbolic graph with oscillating rings to show the conditions are not vacuous. Kondrashov (2025) adds further sufficient parabolicity criteria obtained from differential relations for the coefficients of first-order differential forms. No nontrivial necessary conditions in terms of $f$ are known beyond qualitative ones (for instance, the gradient of $f$ must be unbounded for $M_f$ to be hyperbolic), so an $f$ satisfying none of the sufficient conditions is undecided. Adjacent lines do not close the gap: Del Prete and Gimeno i Garcia (2024) characterize parabolicity completely, but for surfaces invariant under a Killing field (a vector field whose flow preserves the metric, i.e. an infinitesimal symmetry — a class disjoint from general graphs); Bravo and Cortissoz (2026) extend Milnor's rotationally symmetric criterion to biharmonic functions (solutions of the iterated Laplace equation $\Delta^2u=0$), within the symmetric framework.

Problem Statement

Let $f:\mathbb{R}^2\to\mathbb{R}$ be a smooth function and let $M_f=\{(x,y,f(x,y)):(x,y)\in\mathbb{R}^2\}$ be the complete surface with the metric induced from Euclidean $\mathbb{R}^3$. Find an effective criterion for the conformal type of $M_f$ in terms of the function $f$: necessary and sufficient conditions — expressed through the values and derivatives of $f$ or through explicit integral functionals of them, verifiable from $f$ itself rather than from intrinsic objects such as the conformal map, the Green's function, or the capacities of the induced conformal surface, which presuppose knowledge of the type — that decide whether $M_f$ is parabolic or hyperbolic.

This is the graph-surface question posed by Milnor (1977) as the open half of his problem of deciding whether a surface is parabolic or hyperbolic, and canonically restated in Grigor'yan's 1999 survey as an open problem attributed to Milnor: 'Let $M$ be a graph of a smooth function $z=f(x,y)$ defined for all $(x,y)\in\mathbb{R}^2$. Find a criterion for parabolicity of $M$ in terms of the function $f$.' The criterion must apply to all smooth $f$: no symmetry, monotonicity, or boundedness restriction may be imposed, and the known sufficient conditions (capacity and volume tests; the integral tests of Keselman et al. 2024) show what is meant by effective but decide only part of the class. Partial progress — complementary sufficient conditions, necessary conditions matching them, and borderline examples separating them — is genuine progress toward, but does not by itself close, the problem: closure means that the type of every entire graph is decided by conditions checked from $f$.

References

1. John Milnor, "On Deciding Whether a Surface is Parabolic or Hyperbolic", The American Mathematical Monthly 84(1):43-46 (1977). DOI: 10.1080/00029890.1977.11994278, https://doi.org/10.1080/00029890.1977.11994278. (Publisher text paywalled.)
2. Lars V. Ahlfors, "Sur le type d'une surface de Riemann", Comptes Rendus de l'Académie des Sciences Paris 201:30-32 (1935). Zbl: 0011.40701, https://zbmath.org/3017833.
3. Rolf Nevanlinna, "Ein Satz über offene Riemannsche Flächen", Annales Academiae Scientiarum Fennicae, Series A 54(3):1-18 (1940). Zbl: 0024.22104, https://zbmath.org/3039003.
4. Heinz Huber, "Riemannsche Flächen von hyperbolischem Typus im euklidischen Raum", Mathematische Annalen 139(2):140-146 (1959). DOI: 10.1007/bf01354872, https://doi.org/10.1007/bf01354872.
5. James A. Jenkins, "On hyperbolic surfaces in three-dimensional Euclidean space", Michigan Mathematical Journal 8(1):1-5 (1961). DOI: 10.1307/mmj/1028998507, https://doi.org/10.1307/mmj/1028998507.
6. Robert Osserman, "Hyperbolic surfaces of the form z=f(x,y)", Mathematische Annalen 144(1):77-79 (1961). DOI: 10.1007/bf01396545, https://doi.org/10.1007/bf01396545.
7. Reto A. Rüedy, "Embeddings of open Riemann surfaces", Commentarii Mathematici Helvetici 46(1):214-225 (1971). DOI: 10.1007/bf02566840, https://doi.org/10.1007/bf02566840.
8. S. Y. Cheng and S. T. Yau, "Differential equations on riemannian manifolds and their geometric applications", Communications on Pure and Applied Mathematics 28(3):333-354 (1975). DOI: 10.1002/cpa.3160280303, https://doi.org/10.1002/cpa.3160280303.
9. V. M. Miklyukov, "Some criteria for parabolicity and hyperbolicity of the boundary sets of surfaces", Izvestiya: Mathematics 60(4):763-809 (1996). DOI: 10.1070/im1996v060n04abeh000080, https://doi.org/10.1070/IM1996V060N04ABEH000080.
10. Peter G. Doyle, "On deciding whether a surface is parabolic or hyperbolic", Contemporary Mathematics 73:41-48 (1988). DOI: 10.1090/conm/073/954627, https://doi.org/10.1090/conm/073/954627.
11. A. A. Grigor'yan, "On the existence of positive fundamental solutions of the Laplace equation on Riemannian manifolds", Mathematics of the USSR-Sbornik 56(2):349-358 (1987). DOI: 10.1070/sm1987v056n02abeh003040, https://doi.org/10.1070/sm1987v056n02abeh003040.
12. Alexander Grigor’yan, "Analytic and geometric background of recurrence and non-explosion of the Brownian motion on Riemannian manifolds", Bulletin of the American Mathematical Society 36(2):135-249 (1999). DOI: 10.1090/S0273-0979-99-00776-4, https://doi.org/10.1090/S0273-0979-99-00776-4.
13. Vladimir Keselman, Tatyana Igonina, Olga Kozlova, Olga Paraskevopulo, "Conditions of parabolic and hyperbolic types of nonparametric surface", Mathematical Physics and Computer Simulation 27(3):27-37 (2024). DOI: 10.15688/mpcm.jvolsu.2024.3.3, https://doi.org/10.15688/mpcm.jvolsu.2024.3.3. (In Russian.)
14. Alexander Kondrashov, "Some relations that ensure the parabolicity of the type", Mathematical Physics and Computer Simulation 28(4):5-23 (2025). DOI: 10.15688/mpcm.jvolsu.2025.4.1, https://doi.org/10.15688/mpcm.jvolsu.2025.4.1. (In Russian.)
15. Andrea Del Prete, Vicent Gimeno i Garcia, "Parabolicity of Invariant Surfaces", The Journal of Geometric Analysis 34(4), article 109 (2024). DOI: 10.1007/s12220-024-01552-6, https://doi.org/10.1007/s12220-024-01552-6; preprint arXiv:2312.02837, https://arxiv.org/abs/2312.02837.
16. John E. Bravo, Jean C. Cortissoz, "On Milnor's Criterion for Deciding Whether a Surface is Hyperbolic or Parabolic for Biharmonic Functions", The American Mathematical Monthly (2026), online first, 18 pp. DOI: 10.1080/00029890.2026.2699077, https://doi.org/10.1080/00029890.2026.2699077; preprint arXiv:2502.05249, https://arxiv.org/abs/2502.05249.
17. G. Bessa, Vicent Gimeno i Garcia, Leandro Pessoa, Alberto Setti, "On Doyle-Grigor'yan criterion for non-parabolicity", Proceedings of the American Mathematical Society 152(6):2697-2700 (2024). DOI: 10.1090/proc/16804, https://doi.org/10.1090/proc/16804.