Optimal ε–δ–dimension constants in the Freedman–Wang controlled cancellation theorem for controlled modules

Freedman and Wang's chapter "Controlled Linear Algebra" (Prospects in Topology, Princeton, 1995) proves a quantitative controlled cancellation theorem: if $X$ is a Riemannian $n$-manifold and the coefficient group $\pi$ satisfies $\mathrm{Wh}(\pi\times\mathbb{Z}^i)=0$ for $0\le i\le n$, then whenever $\sqrt{n}\cdot 6^{\,n^2+1}\cdot\delta(x)<\varepsilon(x)<\bar{\varepsilon}(x)$ pointwise, every $\delta$-isomorphism between controlled $\mathbb{Z}[\pi]$-modules over $X$ can be deformed, by stabilizations and controlled elementary row operations of spread less than $\varepsilon$, to a geometric map; in the Euclidean case the corresponding local constant improves to $\sqrt{n}\cdot 6^{\,n+1}$. The authors present the chapter as an exposition of Quinn's qualitative vanishing theorem "with perhaps some new estimates on the optimal relations between $\varepsilon$, $\delta$ and dimension", and the published constants are manifestly soft: the manifold constant arises by composing the Euclidean estimate with a $6^{\,(n-1)n}$ loss from a second, big-handle induction. No later work is known to improve, bound from below, or sharpen these constants; the later controlled- and bounded-algebra literature (squeezing theorems, Farrell–Jones methods) develops the qualitative companion principles without explicit constants. The open problem is to determine the optimal relation between $\varepsilon$, $\delta$, and the dimension $n$: improved upper bounds, lower-bound (worst-case) examples, and identification of which ingredients of the proof force the exponential loss.

Background

Controlled topology studies algebraic cancellation problems in which the algebraic data is indexed by the points of a metric space and errors are measured geometrically rather than purely algebraically. The subject grew out of the high-dimensional manifold classification program: obstructions to simplifying homotopy equivalences and $h$-cobordisms (cobordisms that deformation-retract onto both boundary components) live in Whitehead groups, and Quinn's "Ends of Maps" series (1979–1988) showed that these obstructions can be killed when the algebra is controlled near the ends of a map.

The algebraic setting of the chapter is as follows. Let $\pi$ be a group and $\mathbb{Z}[\pi]$ its integral group ring, the ring of finite formal integer combinations of elements of $\pi$ with multiplication induced by the group law. The Whitehead group $\mathrm{Wh}(\pi)$ is the quotient of the group of invertible matrices over $\mathbb{Z}[\pi]$ (stabilized by identity summands) by the subgroups generated by elementary matrices and by trivial units $\pm g$ with $g\in\pi$; it measures the failure of an invertible matrix to be reducible to the identity by elementary row and column operations, and $\mathrm{Wh}(\pi)=0$ means every invertible matrix can be so reduced after stabilization. A controlled module $(M,X,\varphi)$ is a free $\mathbb{Z}[\pi]$-module $M$ with a countable preferred basis $S_M$, together with a proper control map $\varphi:S_M\to X$ into a Riemannian manifold $X$ (proper means that only finitely many basis elements map into any compact subset of $X$); the basis carries the discrete topology. Given a second controlled module $(N,X,\psi)$, the spread of a relation $R$ between the bases at a basis element $e_i$ is $\sup\{d(\varphi(e_i),\psi(f_j)) : f_j\in R(e_i)\}$, where $d$ is the Riemannian distance on $X$. Control is measured pointwise by positive functions $\delta:X\to(0,\infty)$: a homomorphism $\alpha:M\to N$ is a $\delta$-homomorphism if the relation "$f_j$ occurs in $\alpha(e_i)$" has spread less than $\delta$ at every $e_i$, and a $\delta$-isomorphism if it admits a two-sided $\delta$-inverse.

Two kinds of moves are allowed on such maps. A Type I elementary stabilization adjoins one new basis element to each of $S_M$ and $S_N$, with control points closer than $\delta$, and extends $\alpha$ by sending the new element to the new one (algebraically, replacing the matrix $A$ of $\alpha$ by $I\oplus A$); a Type II elementary row operation left-multiplies $A$ by an elementary matrix with diagonal entries in $\pm\pi$, performed with controlled spread. A deformation is a locally finite composition of these moves, and its spread is the spread of the underlying basis-level relation. At the ideal end sit geometric maps: isomorphisms induced by a bijection of the bases that respects the control maps, that is, spread-zero position matchings.

The main theorem of the chapter (Theorem 4.1) states: if $X$ is a Riemannian $n$-manifold and $\pi$ is a group with $\mathrm{Wh}(\pi\times\mathbb{Z}^i)=0$ for all $0\le i\le n$ (a hypothesis called $K$-flatness after Farrell–Jones, who proved it for large classes of groups, among them fundamental groups of closed negatively curved manifolds), then there is a positive function $\bar{\varepsilon}:X\to(0,\infty)$ such that for all positive $\varepsilon,\delta:X\to(0,\infty)$ with

$$\varepsilon(x)<\bar{\varepsilon}(x)\quad\text{and}\quad\sqrt{n}\cdot 6^{\,n^2+1}\cdot\delta(x)<\varepsilon(x)\quad\text{for all }x\in X,$$

every $\delta$-isomorphism between controlled modules over $X$ can be deformed, after stabilization, by an $\varepsilon$-deformation to a geometric map. In the Euclidean case, where the control space is a compact subset of Euclidean $n$-space $E^n$, the corresponding local statement (Lemma 4.2) holds with the better constant $\sqrt{n}\cdot 6^{\,n+1}$.

The constants come from an explicit proof chain. A combinatorial matching step (a Hall-type marriage theorem with a constraint the authors call the "handsome boys" condition) first matches bases and reduces the problem to an automorphism of a single controlled module. That automorphism is then canceled handle by handle, where a handle decomposition is a filtration of the control space by pieces diffeomorphic to discs. This cancellation runs as two inductions: the small-handle induction cancels the automorphism on the low-index handles using the torus trick, and the big-handle induction cancels it over the enlarged big handles. The tools are: the torus trick (precomposing with a covering of a flat torus, so that deck translations, the covering-space symmetries that permute the sheets of the cover, endow the lifted module with the structure of a free $\mathbb{Z}[\pi\times\mathbb{Z}^m]$-module and the hypothesis $\mathrm{Wh}(\pi\times\mathbb{Z}^m)=0$ yields a factorization into elementary matrices); a "ship-in-water" permutation truncation that cuts the infinite lifted automorphism back to a large ball with spread independent of the ball's radii; a tapered retraction controlling the spread at infinity; and Lemma 4.1, which shows that the block-diagonal automorphism $\begin{pmatrix}A&0\\0&A^{-1}\end{pmatrix}$ of a finite-rank controlled module is a $3\delta$-deformation of the identity. Step I of Theorem 4.1 organizes the control space using a Whitney triangulation (a triangulation whose simplices are uniformly close to flat) and a Nash isometric embedding (an embedding of a Riemannian manifold into a Euclidean space that preserves lengths of curves), producing big handles that are $(1+0.001)$-quasi-isometric (distance-distorting by at most that factor) to subsets of Euclidean space. The manifold constant is obtained by composing the Euclidean constant $\sqrt{n}\cdot 6^{\,n+1}$ with a factor $6^{\,(n-1)n}$ produced by a second induction over these big handles.

Two qualitative principles frame the chapter. The first, the vanishing theorem, proved by Quinn in Section 8 of "Ends of Maps I" and re-proved here with explicit constants, says that Whitehead-type obstructions vanish when only simply connected directions are uncontrolled. The second, the squeezing principle, not covered in the chapter, says that once a threshold amount of control is achieved, any finer control can be achieved; the chapter defers it, and the applications of the theory outside topology (notably to the truncation of wavelet matrices in signal processing), to other work. Later literature developed both principles qualitatively: Pedersen's survey of controlled algebraic $K$-theory (2000); Bartels's squeezing theorems for higher algebraic $K$-theory (2003); the foliated squeezing theorem for geometric modules and the controlled-algebra approach to the Farrell–Jones isomorphism conjecture of Bartels–Farrell–Jones–Reich (2003, 2004); the computation for the infinite dihedral group by Davis–Quinn–Reich (2011); the subdivision theorems for simplicially controlled categories of Adams-Florou (2014); the survey of Reich–Varisco (2018); and the explicit examples of vanishing and squeezing in $K_1$ by Ellis–Rodríguez Cirone–Tartaglia–Vega (2019). All of these frameworks are qualitative: none tracks an explicit dimension-dependent relation between $\varepsilon$ and $\delta$ of the kind Theorem 4.1 provides.

Problem Statement

Determine the optimal relation between the control scales $\varepsilon$ and $\delta$ and the dimension $n$ in the Freedman-Wang controlled cancellation theorem, within the same framework and under the same hypotheses as the chapter (controlled $\mathbb{Z}[\pi]$-modules over a Riemannian control space; deformations given by locally finite Type I stabilizations and Type II elementary row operations; geometric maps as targets; coefficient group $\pi$ with $\mathrm{Wh}(\pi\times\mathbb{Z}^i)=0$ for $0\le i\le n$). Concretely:

1. Upper bounds. Does a function $C(n)$ with strictly better growth than $\sqrt{n}\cdot 6^{\,n^2+1}$ suffice in Theorem 4.1? In particular, can the factor $6^{\,(n-1)n}$ contributed by the big-handle induction be reduced or removed, and can the base 6 in the Euclidean constant $\sqrt{n}\cdot 6^{\,n+1}$ of Lemma 4.2 be improved? In this framework the qualitative existence of a finite constant is already known; the open problem is the quantitative growth of $C(n)$.

2. Lower bounds and sharpness. Do there exist, for infinitely many dimensions $n$, controlled modules and $\delta$-isomorphisms over an $n$-dimensional control space, together with an explicit function $f(n)$, such that no $\varepsilon$-deformation of the given $\delta$-isomorphism to a geometric map exists whenever $\varepsilon(x)<f(n)\cdot\delta(x)$ pointwise? Such examples would settle whether the linear-in-$\delta$ form and/or the dimension-dependence can be improved in general; no lower-bound examples for this theorem exist in the literature.

3. Attribution of the loss. Identify which ingredients of the published proof contribute essential, unavoidable spread loss and which are artifacts of the exposition: the marriage-theorem basis matching, the torus-trick lifting, the ship-in-water permutation truncation, the tapered retraction, the $3\delta$ cancellation of Lemma 4.1, the small-handle induction, and the $(1+0.001)$-quasi-isometric big-handle decomposition obtained via Whitney triangulation and Nash embedding. This includes deciding whether a triangulation-based organization of the argument yields better constants.

The optimal order of the linear constant $C(n)$ is established when a best-possible upper bound is matched by worst-case examples up to explicit absolute factors, that is, when an upper bound $C(n)$ is matched by lower-bound examples whose explicit function $f(n)$ has the same growth order, matched up to explicit absolute factors. Items 1 and 2 are the two components of this relation, and item 3 is a separate attribution question.

References

1. Michael H. Freedman and Zhenghan Wang, "Controlled Linear Algebra", in: Frank Quinn (ed.), Prospects in Topology (Annals of Mathematics Studies 138), Princeton University Press, Princeton, NJ, 1995, pp. 138-156. DOI 10.1515/9781400882588-011, https://doi.org/10.1515/9781400882588-011
2. Frank Quinn, "Ends of maps. I", Annals of Mathematics (2) 110 (1979), 275-331. DOI 10.2307/1971262, https://doi.org/10.2307/1971262
3. Frank Quinn, "Ends of maps. II", Inventiones Mathematicae 68 (1982), 353-424. DOI 10.1007/BF01389410, https://doi.org/10.1007/BF01389410
4. F. T. Farrell and L. E. Jones, "K-theory and dynamics. I", Annals of Mathematics (2) 124 (1986), 531-569. DOI 10.2307/2007092, https://doi.org/10.2307/2007092
5. Michael H. Freedman and William H. Press, "Truncation of wavelet matrices: Edge effects and the reduction of topological control", Linear Algebra and its Applications 234 (1996), 1-19. DOI 10.1016/0024-3795(94)00039-5, https://doi.org/10.1016/0024-3795(94)00039-5
6. Erik Kjaer Pedersen, "Controlled algebraic K-theory, a survey", Contemporary Mathematics 258 (2000), 351-368. DOI 10.1090/conm/258/1778117, https://doi.org/10.1090/conm/258/1778117
7. Christopher W. Stark, "Topological Rigidity Theorems", in: Handbook of Geometric Topology, North-Holland/Elsevier, Amsterdam, 2001, pp. 1045-1084. DOI 10.1016/B978-044482432-5/50021-4, https://doi.org/10.1016/B978-044482432-5/50021-4
8. Arthur C. Bartels, "Squeezing and Higher Algebraic K-Theory", K-Theory 28 (2003), 19-37. DOI 10.1023/A:1024166521174, https://doi.org/10.1023/A:1024166521174
9. Arthur Bartels, Tom Farrell, Lowell Jones and Holger Reich, "A foliated squeezing theorem for geometric modules", in: High-Dimensional Manifold Topology (Proceedings of the school, ICTP, Trieste, May 21 - June 8, 2001), World Scientific, River Edge, NJ, 2003, pp. 1-21. DOI 10.1142/9789812704443_0001, https://doi.org/10.1142/9789812704443_0001, ISBN 981-238-223-2
10. Arthur Bartels, Tom Farrell, Lowell Jones and Holger Reich, "On the Isomorphism Conjecture in algebraic K-theory", Topology 43 (2004), 157-213. DOI 10.1016/S0040-9383(03)00032-6, https://doi.org/10.1016/S0040-9383(03)00032-6
11. James F. Davis, Frank Quinn and Holger Reich, "Algebraic K-theory over the infinite dihedral group: a controlled topology approach", Journal of Topology 4 (2011), 505-528. DOI 10.1112/jtopol/jtr009, https://doi.org/10.1112/jtopol/jtr009
12. Spiros Adams-Florou, "Algebraic subdivision in simplicially controlled categories", arXiv:1405.2973 (2014), https://arxiv.org/abs/1405.2973
13. Holger Reich and Marco Varisco, "Algebraic K-theory, assembly maps, controlled algebra, and trace methods", in: Space - Time - Matter, De Gruyter, Berlin, 2018, pp. 1-50. DOI 10.1515/9783110452150-001, https://doi.org/10.1515/9783110452150-001
14. Eugenia Ellis, Emanuel Rodriguez Cirone, Gisela Tartaglia and Santiago Vega, "Two examples of vanishing and squeezing in $K_1$", arXiv:1907.06135 (2019), https://arxiv.org/abs/1907.06135
15. Michael H. Freedman, "Controlled Mather-Thurston theorems", EMS Surveys in Mathematical Sciences 9 (2022), 415-445. DOI 10.4171/EMSS/63, https://doi.org/10.4171/EMSS/63