---
title: Sublinear Bilipschitz Retraction
url: https://www.emergentmind.com/topics/sublinear-bilipschitz-retraction
type: topic
---

# Sublinear Bilipschitz Retraction

Sublinear bilipschitz retraction is a coarse-geometric notion in which a map back onto a subspace, quotient, or model space is required to respect distances up to bounded multiplicative distortion and an additive error that is sublinear in the distance from a basepoint. In the framework of sublinear biLipschitz equivalence (SBE), the error is allowed to grow unboundedly, but strictly slower than linearly at infinity [2301.05086]. A related question, stated for separated nets in Euclidean space, asks: given a separated net in \( \mathbb{R}^d \), can one construct a bilipschitz map from the ambient space back onto the net with distortion growing sublinearly at infinity; this is explicitly distinguished from the extension problem, whose focus is extending an embedding from the net to the whole space rather than retracting the ambient space onto the net [2410.22294].

## 1. Formal framework

A function \( u:\mathbb{R}_{\geq 0} \to \mathbb{R} \) is called sublinear if it is admissible—non-decreasing, of sublinear growth \( \lim_{r\to\infty} \frac{u(r)}{r} = 0 \), and doubling—and is also subadditive: \( u(s+t) \leq u(s) + u(t) \) for all \( s, t \) [2301.05086]. Given pointed metric spaces \((X, d_X, x_0)\), \((Y, d_Y, y_0)\), an \((L, u)\)-SBE is a map \(f: X \to Y\) such that \( f(x_0) = y_0 \), such that for all \( x_1, x_2 \in X \),
\[
\frac{1}{L}|x_1 - x_2|_X - u\big(|x_1|_X \vee |x_2|_X\big) \leq |f(x_1) - f(x_2)|_Y \leq L|x_1 - x_2|_X + u\big(|x_1|_X \vee |x_2|_X\big),
\]
and such that for all \( y \in Y \), there exists \( x \in X \) with \( |y - f(x)|_Y \leq u(|y|_Y) \) [2301.05086]. Every quasi-isometry is an SBE, but SBE is coarser, allowing more “wild” metric behaviors far from the basepoint [2301.05086].

For retracts, one formulation is explicit: a sublinearly Lipschitz retract (\(o(r)\)-retract) is a \( o(r) \)-Lipschitz map \( g: Y\to X \) for which there exists \( f: X \to Y \), with \( g \circ f \) \( o(r) \)-close to identity on \( X \) [1702.06618]. A second formulation, used for solvable Lie groups, says that the notion of a sublinear bilipschitz retraction refers to the existence of a map that is a sublinear bilipschitz embedding from a group \(H\) into \(G\), with a coarse sublinear inverse; the map is not required to be an actual group homomorphism, but rather to respect the metric up to sublinear terms [2410.05042]. These formulations are compatible at the level of large-scale geometry: both encode a retraction phenomenon modulo sublinear error.

A nearby one-sided notion is sublinear covering. For a function \( u:\mathbb{R}_{\geq 0}\to \mathbb{R} \) and sets \( Y, Z\subset X \), one defines
\[
N_u(Y) := \{x \in X : d(x, Y) \leq u(|x|)\},
\]
and says that \(Y\) sublinearly covers \(Z\) if \( Z \subset N_u(Y) \) [2301.05086]. This notion is weaker than a retract, but it plays a central role in rigidity statements adjacent to retraction theory.

## 2. Canonical constructions in Lie groups

The most concrete positive example in the supplied literature comes from solvable Lie groups. Let \(G\) be a completely solvable Lie group, let \(N = \operatorname{R}_{\exp} G\) be the exponential radical, and let \(H = G/N\). Then the projection \( \pi: G \rightarrow H \) is \( O(\log) \)-Lipschitz, and there exists a map \( f: H \to G \), a cross-section, which is \( O(\log) \)-Lipschitz, such that \( \pi \circ f \) is \( O(\log) \)-close to the identity on \(H\). Moreover, for any \(X\) in a Cartan subalgebra, \(f \circ \pi (\exp{tX}) = \exp(tX)\) for all \(t\). This means \(H\) is an \( O(\log) \)-bilipschitz retract of \(G\) [2410.05042].

This construction depends on the algebraic structure of the ambient group. The supplied description states that the construction relies heavily on the algebraic splitting conditions in completely solvable groups, especially those where the extension of the exponential radical is well behaved [2410.05042]. A closely related statement appears in the earlier SBE literature: if \(G\) is a simply connected Lie group with exponential radical \(E\), then the quotient \(G/E\) is an SBE-retract of \(G\) [1702.06618]. In both cases, the quotient by the exponential radical serves as the retract, and the distortion is controlled by a sublinear function.

These examples are significant because they provide actual coarse retraction mechanisms rather than only equivalence results. They also show that logarithmic error terms can arise naturally in non-nilpotent solvable settings. A plausible implication is that the algebraic decomposition \(G \to G/\operatorname{R}_{\exp} G\) is one of the principal settings in which positive sublinear bilipschitz retraction results are presently available in explicit form.

## 3. Rigidity phenomena and obstructions

Sublinear retractions are constrained by the same rigidity mechanisms that govern SBE. For lattices in semisimple Lie groups, the class of lattices in groups that do not admit \( \mathbb{R} \)-rank \( 1 \) factors is SBE complete: if \( \Lambda \) is SBE to an irreducible lattice \( \Gamma\leq G \), then \( \Lambda \) can be homomorphically mapped into \(G\) with finite kernel and image a lattice in \(G\), and the uniformity matches between \( \Gamma \) and \( \Lambda \) [2301.05086]. In the one-sided setting, if \( \Lambda\leq G \) is a discrete subgroup that sublinearly covers a lattice \( \Gamma\leq G \), then \( \Lambda \) is itself a lattice, under the stated rank and irreducibility assumptions [2301.05086]. The same source also states a necessary condition for sublinear covering leading to a lattice: the covering error \(u\) must be sublinear; if it were linear, \( u(r) = cr \), the conclusion could fail.

Boundary invariants provide further obstructions. For proper geodesic metric spaces, if \( \Phi:X\to Y \) is an \((L,\theta)\)-SBE and \( \kappa \) dominates \( \theta \), then the induced map on \( \kappa \)-Morse boundaries is a homeomorphism [2211.01023]. The same account states that the existence of a sublinear retraction \(X\to Y\), that is, a coarse “sublinear closest point projection,” is obstructed if the induced map would force incompatible Morse boundaries; if one space’s boundary is connected, and the other's is totally disconnected, no SBE retraction can exist [2211.01023]. This gives a direct topological obstruction to the existence of coarse retracts.

A parallel rigidity statement is given for negatively curved homogeneous spaces. In the Heintze group context, the only way for an SBE retraction to exist between a Heintze group \(S\) and its diagonal model \(S_\infty\) is if the groups are isomorphic; there are no non-trivial SBE retractions between non-isomorphic spaces of this type [1905.08981]. The same source ties this to invariance of the sublinear conformal dimension and of function spaces of locally bounded \( p \)-variation. Together with the boundary-dimension constraints for hyperbolic spaces, this shows that sublinear bilipschitz retraction remains a rigid notion even though SBE is weaker than quasi-isometry [1801.05163].

## 4. Relation to bilipschitz extension theory

The retraction problem is repeatedly separated from extension theory. In the planar separated-net paper, the sublinear bilipschitz retraction is described as a related but different question: given a separated net in \( \mathbb{R}^d \), can one construct a bilipschitz map from the ambient space back onto the net, a “retraction,” with distortion growing sublinearly at infinity [2410.22294]. The same paper does not address that problem directly. Instead, it proves that every \(L\)-bilipschitz mapping \( \mathbb{Z}^2\to\mathbb{R}^2 \) can be extended to a \( C(L) \)-bilipschitz mapping \( \mathbb{R}^2\to\mathbb{R}^2 \), and extends this to every separated net in \( \mathbb{R}^2 \) [2410.22294]. The authors state that their techniques may be related to the retraction question, but the focus is strictly on extending the embedding from the net to the whole space, rather than retraction.

A second contrast comes from Kirszbraun-type extension theory. The explicit formula
\[
\widetilde{G}(x) := \nabla_Y(\mathrm{conv}(g))(x, 0)
\]
constructs a global extension preserving the Lipschitz constant, and in the strongly biLipschitz setting preserves the strong biLipschitz constant as well [1810.10288]. However, the same source is explicit that the construction is not a retraction onto the original subset, does not yield a sublinear bilipschitz retraction, and does not imply the existence of such retractions in Hilbert space [1810.10288]. It further describes the existence of such retractions as a much deeper and in general open/difficult problem.

The distinction is structural. Extension theorems enlarge the domain of a prescribed map while maintaining metric control. Retraction theorems, by contrast, impose an ambient-to-subspace or ambient-to-quotient mechanism, and therefore interact with coarse surjectivity, boundary invariants, and algebraic rigidity in ways that extension theory does not.

## 5. Retractions between finite subset spaces

A different retraction problem appears in finite subset spaces \(X(n)\), where sublinearity refers not to an additive error in space, but to the growth of the Lipschitz constants \( \operatorname{Lip}(r_n) \) as \( n \to \infty \). For normed spaces or Hadamard spaces, this growth is provably not sublinear. If \(X\) is either a real normed space with \( \dim X \geq 1 \), or a Hadamard space containing more than one point, and \( r: X(n) \to X(n-1) \) is a Lipschitz retraction, then
\[
\operatorname{Lip}(r) \ge n - 3,
\]
so no uniformly-Lipschitz \( O(1) \) or sublinear \( o(n) \) growth is possible [2005.13579]. The same paper states that even in Hilbert spaces, \( \operatorname{Lip}(r_n) \gtrsim n \) is necessary.

At the same time, ordinary Lipschitz retractions do exist in several low-step or specialized cases. In any normed space, there exists a \(1\)-Lipschitz retraction \( X(2)\to X \), given by the average map, and there exists a Lipschitz retraction \( X(3)\to X(2) \) with explicit constant \(731\) [1811.00603]. For every \( n\ge 1 \), there are Lipschitz retractions \( X(n)\to X \) and \( X(n)\to X(2) \), and for every \( n\ge2 \) there exists a retraction \( r:X(n)\to X(n-1) \) which is Hölder-continuous on bounded sets [1811.00603]. The same source also states that no bilipschitz or Lipschitz retraction is constructed except for \( n=2,3 \), and that no sublinear Lipschitz constant retractions \( X(n)\to X(n-1) \) are established.

This finite-subset-space literature is not an instance of SBE retraction in the sense of coarse geometry. Nevertheless, it supplies an important cautionary comparison: even when retractions exist, quantitative control can be forced to grow linearly, and not sublinearly, with the complexity parameter \(n\).

## 6. Scope, variants, and open directions

The supplied literature leaves several aspects of sublinear bilipschitz retraction open. In the lattice setting, open questions include the existence of SBE-equivalent but not quasi-isometric groups, the behavior with \( \mathbb{R} \)-rank one factors or products of \( \mathrm{SL}_2 \) factors, and explicit examples of sublinear but not bounded covering [2301.05086]. In finite subset spaces, the existence of Lipschitz retractions \( X(n)\to X(n-1) \) for general Banach spaces remains unresolved for \( n>3 \), except for Hilbert and CAT(0) spaces [1811.00603].

For separated nets in \( \mathbb{R}^d \), the retraction question remains separate from the now-established planar extension theorem. The planar result gives the first positive result in the plane for bilipschitz extension from \( \mathbb{Z}^2 \) and from arbitrary separated nets in \( \mathbb{R}^2 \), with explicit double-exponential upper bounds on the extension constant, but it does not settle the retraction problem [2410.22294]. The same source states that the developed machinery could inspire similar approaches for related problems in metric geometry, and for studying retractions or more general extension phenomena.

Accordingly, sublinear bilipschitz retraction is best understood not as a single theorem, but as a family of coarse retraction phenomena. In solvable Lie groups it is realized concretely through \( O(\log) \)-bilipschitz retracts [2410.05042]. In hyperbolic, Heintze, and higher-rank settings it is strongly constrained by boundary structure, conformal dimension, Morse-boundary topology, and lattice rigidity [2211.01023] [1905.08981] [2301.05086]. In Euclidean net geometry and in finite subset spaces, neighboring extension and retraction problems show that positive existence statements and sharp quantitative obstructions coexist, but do so in sharply different forms [2410.22294] [2005.13579].

Source: https://www.emergentmind.com/topics/sublinear-bilipschitz-retraction