---
title: Affine Maximal Maps
url: https://www.emergentmind.com/topics/affine-maximal-maps
type: topic
---

# Affine Maximal Maps

Searching arXiv for recent and foundational papers on affine maximal maps and related usages.
“Affine maximal maps” denotes several related but non-identical extremal notions. In the most direct affine differential-geometric usage, it refers to affine maximal surfaces with singularities in unimodular affine \(3\)-space, defined by Aledo–Martínez–Milán through holomorphic Weierstrass data and an affine metric that is allowed to vanish. Closely related literatures use “affine maximal type hypersurfaces” for convex graphs solving a fourth-order Monge–Ampère type equation, “Calabi affine maximal surfaces” for Calabi hypersurfaces satisfying \(\Delta\ln\det(f_{ij})=0\), and “fully affine maximal curves” for extremals of fully affine arclength. This suggests that the expression is best understood as an umbrella term for affine-invariant extremal objects, with distinct variational functionals, completeness notions, and Bernstein-type rigidity statements.

## 1. Equiaffine maximality and graph formulations

In the unimodular affine \(3\)-space \(\mathbb{R}^3\), a transversal vector field \(\xi\) along an immersion \(\psi:M\to\mathbb{R}^3\) determines a torsion-free affine connection \(\nabla\), a symmetric quadratic form \(h\), a shape operator \(S\), and a \(1\)-form \(\tau\) through the Gauss–Weingarten equations
\[
\begin{cases}
D_X d\psi(Y) = d\psi(\nabla_XY) + h(X,Y)\,\xi, \\
D_X\xi = -d\psi(SX) + \tau(X)\,\xi .
\end{cases}
\]
If \(h\) is positive definite, there is a unique affine normal \(\xi\) for which \((\psi,\xi)\) is a Blaschke immersion, and the conormal map \(N\) is defined by
\[
\langle N, d\psi(X)\rangle = 0, \qquad \langle N,\xi\rangle = 1 .
\]
The affine mean curvature is \(\frac12\operatorname{tr}S\). A Blaschke immersion is **affine maximal** if the affine mean curvature vanishes everywhere; equivalently,
\[
\Delta_h N = 0 .
\]
An **improper affine sphere** is characterized by \(S=0\), and every improper affine sphere is affine maximal [2507.10035].

For locally strongly convex graphs, affine maximality becomes a fourth-order PDE. If \(\psi\) is written locally as the graph of a function \(u(x,y)\) with positive definite Hessian, then affine maximality is equivalent to
\[
\rho_{yy}\rho_{xx}-2\rho_{xy}^2+\rho_{xx}\rho_{yy}=0,
\qquad
\rho := (\det(\mathrm{Hess}\,u))^{-3/4}.
\]
In higher dimensions, the affine maximal type equation studied by Trudinger–Wang and later work is
\[
U^{ij} D_{ij}w = 0 \quad\text{in }\Omega,
\qquad
w := (\det D^2u)^{-\theta},
\]
for a strictly convex \(C^4\) potential \(u\). In the classical affine maximal case,
\[
\theta = \frac{N+2}{N+1},
\]
the affine metric is
\[
A_{ij} = (\det D^2u)^{-1/(N+2)} u_{ij},
\]
the affine mean curvature is
\[
H = (\det D^2u)^{-1/(N+2)},
\]
and the equation becomes
\[
\Delta_M H = 0 .
\]
Thus the classical affine maximal equation appears as one distinguished member of a \(\theta\)-family of affine maximal type equations [2103.08921].

## 2. Holomorphic representation and affine maximal maps with singularities

For a simply connected locally strongly convex affine maximal immersion, there exists a holomorphic map
\[
\Phi:M\to\mathbb{C}^3
\]
such that
\[
N = \Phi+\bar\Phi,
\qquad
h = -2i\,[\Phi+\bar\Phi,\ \partial\Phi,\ \bar\partial\Phi],
\]
and
\[
\psi = 2\Re\left(i\int (\Phi+\bar\Phi)\times d\Phi\right)
      = -i\,\Phi\times\bar\Phi - 2\Im\int \Phi\times d\Phi .
\]
Conversely, if \(\Phi\) is holomorphic and satisfies the single-valuedness, period, and positivity conditions stated in the representation theorem, the Lelieuvre formula defines a locally strongly convex affine maximal immersion [2507.10035].

Aledo–Martínez–Milán extend this to surfaces with admissible singularities. A map \(\psi:M\to\mathbb{R}^3\) is an **affine maximal map** if there exists a holomorphic map \(\Phi\) on the universal cover \(\widetilde M\) such that \([\Phi+\bar\Phi,\partial\Phi,\bar\partial\Phi]\) is not identically zero and \(\psi\) is given locally by the Lelieuvre formula. In this setting, \(\Phi\) is the **Weierstrass data**, \(N:=\Phi+\bar\Phi\) is the **conormal map**, and
\[
h:=-2i[\Phi+\bar\Phi,\partial\Phi,\bar\partial\Phi]
\]
is the **affine metric**. Points where \(h\) vanishes are singular points.

Because \(h\) may degenerate, completeness is modified accordingly. An affine maximal map is **complete** if there exists a symmetric \((0,2)\)-tensor \(T\) with compact support such that \(|h|+T\) is a complete Riemannian metric. For complete affine maximal maps,
\[
M \cong \widehat{M}\setminus\{p_1,\dots,p_n\}, \quad n\ge 1,
\]
where \(\widehat M\) is a compact Riemann surface and the punctures are ends. An end is **regular** if \(d\Phi\) extends meromorphically to the puncture. In this framework, a complete regular affine maximal map with one embedded end is, equiaffinely, the elliptic paraboloid; this is the extended affine Bernstein theorem [2507.10035].

## 3. Affine maxfaces and the minimal-surface correspondence

The paper “A class of affine maximal surfaces with singularities and its relationship with minimal surface theory” isolates a special subclass of affine maximal maps, called **affine maxfaces**, by imposing that the conormal map \(N\) be a Euclidean conformal minimal immersion [2507.10035]. On a simply connected Riemann surface, \(N\) then has the classical minimal Weierstrass representation
\[
N = \Re\int (1-g^2,\ i(1+g^2),\ 2g)\,\omega,
\qquad
\Phi = \frac12\int (1-g^2,\ i(1+g^2),\ 2g)\,\omega,
\]
where \(g\) is meromorphic and \(\omega\) is holomorphic. The Euclidean first fundamental form of \(N\) is
\[
d\sigma^2 = (1+|g|^2)^2 |\omega|^2,
\]
and the Euclidean unit normal is
\[
\nu
= \left( \frac{2\Re g}{1+|g|^2},\ 
 \frac{2\Im g}{1+|g|^2},\ 
 \frac{-1+|g|^2}{1+|g|^2} \right)
= \Pi\circ g .
\]
In affine language, \(\nu\) is the affine Gauss map and \(g\) is its meromorphic model.

The affine metric is
\[
h = \langle N,\nu\rangle\, d\sigma^2,
\]
so the singular set is
\[
\mathcal{S} = \{p\in M;\ \langle N(p),\nu(p)\rangle = 0\}.
\]
This immediately yields a second completeness notion: an affine maxface is **weakly complete** if the minimal metric \(d\sigma^2\) is complete. Completeness in the affine sense implies weak completeness. More precisely, an affine maxface is complete and regular if and only if it is weakly complete, of finite total curvature, and its singular set is compact. Here finite total curvature means finite total curvature of the minimal conormal immersion,
\[
\int_M K_{d\sigma^2}\,dA_{d\sigma^2}
= -\int_M \frac{2i\,dg\wedge d\bar g}{(1+|g|^2)^2}
= -4\pi \deg(g).
\]

The minimal-surface correspondence imports the full Osserman theory. If
\[
\psi:\widehat{M}_\gamma\setminus\{p_1,\dots,p_n\}\to\mathbb{R}^3
\]
is a complete regular affine maxface, then
\[
-\frac{1}{2\pi}\int_M K_{d\sigma^2}\,dA_{d\sigma^2}
= 2\,\deg(g)
\;\ge\;
-\chi(\widehat{M}_\gamma) + 2n,
\]
with equality if and only if all ends are embedded. The same inequality also holds for weakly complete affine maxfaces of finite total curvature. The class is sharply separated from improper affine fronts: an affine maxface is an improper affine front if and only if its image is contained in an elliptic paraboloid, and any complete affine maxface with constant affine Gauss map is the elliptic paraboloid. The paper further proves that affine maxfaces are fronts and gives explicit criteria for cuspidal edges and swallowtails in terms of \((g,\omega)\), \(N\), and the singular equation \(\langle N,\nu\rangle=0\). Enneper-, catenoid-, helicoid-, Möbius-strip-, and Miyaoka–Sato-type constructions show that weak completeness is abundant, whereas full completeness is obstructed by the behavior of the singular set at the ends [2507.10035].

## 4. Improper affine fronts and value-distribution rigidity

A complementary line of work studies affine Gauss-type maps by Nevanlinna-theoretic methods. The paper “On the maximal number of exceptional values of Gauss maps for various classes of surfaces” develops a general curvature estimate for conformal metrics of the form
\[
ds^2 = (1 + |g|^2)^m \, |\omega|^2
\]
on an open Riemann surface, with \(g\) meromorphic and \(\omega\) holomorphic. The Gaussian curvature is
\[
K_{ds^2}
= -\frac{2m |g_z|^2}{(1+|g|^2)^{m+2} |\hat\omega_z|^2}.
\]
If \(g\) omits \(q\ge m+3\) distinct values in \(\mathbb{CP}^1\), then there exists \(C>0\) such that
\[
\sqrt{|K_{ds^2}(p)|}\le \frac{C}{d(p)},
\]
where \(d(p)\) is the geodesic distance to the boundary. If \(ds^2\) is complete and \(g\) is nonconstant, then \(g\) can omit at most \(m+2\) distinct values. The bound is sharp [1205.4782].

For improper affine fronts in \(\mathbb{A}^3\), Martínez’ representation uses holomorphic functions \(F,G\) with \(\Re(F\,dG)\) exact and \(|dF|+|dG|\) positive definite, and defines
\[
\psi
= \left( G + F,\; |G|^2 - |F|^2 + 2\,\Re\int^{z} \big(G\,dF - F\,dG\big) \right).
\]
The induced metric on the associated special Lagrangian immersion is
\[
d\tau^2 = 2\left(|dF|^2 + |dG|^2\right)
        = 2(1+|\nu|^2)\,|dG|^2,
\qquad
\nu := \frac{dF}{dG}.
\]
Here \(\nu\) is the **Lagrangian Gauss map**, and **weak completeness** means completeness of \(d\tau^2\). Since this is the case \(m=1\), the general theorem yields a sharp Picard-type bound: if the Lagrangian Gauss map of a weakly complete improper affine front is nonconstant, then it can omit at most \(3\) values. This gives a short proof of the affine Bernstein theorem: any affine complete improper affine sphere in \(\mathbb{A}^3\) must be an elliptic paraboloid, because nonsingularity implies \(|\nu|=1\) is omitted and therefore \(\nu\) must be constant [1205.4782].

## 5. Higher-dimensional affine maximal type, Calabi affine maximality, and centroaffine extremals

The higher-dimensional Bernstein problem for affine maximal type hypersurfaces asks whether every locally uniformly convex, Euclidean complete affine maximal type \(C^4\)-hypersurface in \(\mathbb{R}^{N+1}\) must be an elliptic paraboloid. In the graph formulation, this asks whether every convex \(C^4\)-solution of
\[
U^{ij} D_{ij}\big[(\det D^2u)^{-\theta}\big] = 0
\]
with Euclidean complete graph must be quadratic. Before 2021, the full Bernstein theorem was known for \(N=1\), all \(\theta>0\), and for \(N=2\), \(\theta\in[3/4,1]\). The paper “Bernstein Problem of Affine Maximal Type Hypersurfaces on Dimension \(N\ge 3\)” shows that for every \(N\ge 3\) and
\[
\theta \in \left(\frac12,\frac{N-1}{N}\right),
\]
there exists a convex domain \(\Omega\subset\mathbb{R}^N\) and a non-quadratic \(C^4\) solution \(u\) of the affine maximal type equation such that
\[
\lim_{x\to\partial\Omega} u(x) = +\infty.
\]
This boundary behavior enforces Euclidean completeness. Hence, in that parameter range, the full Bernstein theorem in the Trudinger–Wang sense fails for \(N\ge 3\). At the same time, the paper proves a radial Bernstein theorem: for \(N\ge 3\) and \(\theta>0\), the only radial solutions satisfying the convexity and regularity conditions are quadratic [2103.08921].

A different affine normalization leads to **Calabi affine maximal surfaces**. For a strictly convex graph \(x_{n+1}=f(x_1,\dots,x_n)\), the Calabi metric is
\[
G = \sum_{i,j} f_{ij}\,dx_i\,dx_j,
\qquad
A_{ijk} = -\tfrac12 f_{ijk}.
\]
Gao’s first variation formula gives the Euler–Lagrange equation
\[
\Delta \ln\det(f_{ij}) = 0,
\]
equivalently \(\operatorname{tr}\mathcal{T}=0\) for the Tchebychev field. In dimension \(2\), the second variation reduces to
\[
V''(0) = -\tfrac12\bigl[(L\varphi,L\varphi) + (\hat{A}\varphi,\hat{A}\varphi)\bigr],
\]
so every Calabi extremal surface is also maximal in the Calabi affine geometry. The paper therefore defines such objects as **Calabi affine maximal surfaces**. It further gives local classifications under conditions such as constant \(|T|\) and flatness with \(|\hat\nabla T|=c>0\), constructs a new complete flat Calabi affine maximal surface, and uses a type II Calabi product to build complete centroaffine extremal hypersurfaces. Notably, the complete centroaffine extremal hypersurfaces established there answer all five centroaffine Bernstein problems posed by Li–Li–Simon in 2004 [2601.10125].

## 6. Fully affine, analytic, and group-theoretic extensions of the phrase

In fully affine plane geometry, Yang considers the fully affine group
\[
GA(2)=GL(2)\ltimes\mathbb{R}^2
\]
and the fully affine length functional
\[
\mathcal{L}[C]=\int_C d\xi .
\]
A curve is **fully affine extremal** if the first variation vanishes, and **fully affine maximal** if, in addition, the second variation is non-positive. The Euler–Lagrange equation is
\[
\varphi_{\xi\xi\xi} + \varphi\,\varphi_{\xi\xi}
+ \frac{2\varphi^2 + 3\varphi + \epsilon}{9}\,\varphi_\xi = 0.
\]
Within this theory, fully affine maximal curves are “much more abundant” and include the explicit curves
\[
y=x^\alpha \qquad
\left(\alpha\notin\{0,1,\tfrac12,2\}\right)
\]
and
\[
y=x\log x.
\]
The same paper classifies solitons of the fully affine heat flow and proves that the only closed solitons are ellipses; moreover, a closed embedded curve converges to an ellipse under the fully affine heat flow [2202.01992].

The phrase also appears in analytic and algebraic contexts that are not part of affine differential geometry proper. In Banach space Lipschitz theory, “maximal affine functions” are affine approximants \(g(t)=at+b\) satisfying
\[
(g) > (f|_B)-\varepsilon .
\]
The pair \((\mathbb{R},\mathbb{R})\) has the maximal approximation by affine property, but fails the maximal uniform approximation by affine property [2410.16607]. In permutation-group theory, \(\mathrm{AGL}_n(\mathbb{Q})\) is a maximal-closed subgroup of \(S_\omega\) for \(2\le n\le\omega\), so affine maps are maximal among closed permutation symmetries of countable rational affine space [1310.8157]. In the affine–additive group \(AA\), linear and radial stretch maps minimize the mean quasiconformal distortion functional; the linear stretch maps also minimize maximal distortion in the corresponding boundary classes [2411.13129]. This suggests that “affine maximal maps” does not designate a single theory, but a family of extremal affine phenomena unified by maximality of area, length, omitted-value behavior, or distortion.

Source: https://www.emergentmind.com/topics/affine-maximal-maps