Independent scholarship · Open accessThe monthly digest

Indexed text attachment

differential-topology-in-higher-categories.tex

Source for Differential Topology in Higher Categories

View paper · Download original file

Lines 5539–5638 of 6119. Source text is displayed literally.

5539by the homotopy allows the base point of \(d\circ e\) to be compared5540with the starting point in \(X\). With that comparison in place,5541the map on each homotopy group induced by \(e\) has a left5542inverse induced by \(d\). See \cite{hatcher-at} for the cohomology and5543base-point conventions.5544 5545For a closed connected oriented \(n\)-manifold \(X\), with \(n>0\),5546a degree-one reconstruction of \(X\) induces a nonzero map on5547\(H^n(X;\mathbb Z)\). It cannot factor through a manifold of5548dimension less than \(n\), since that manifold has no cohomology5549in degree \(n\). Here one can use a triangulation of the5550intermediate smooth manifold and its dimension bound. This is a5551homotopy obstruction, so it also applies when the reconstruction5552has critical points. A factorization through a contractible5553\(L\) makes the composite nullhomotopic by applying \(d\) to a5554contraction after \(e\). Positive dimension allows degree to5555distinguish this case from the identity.5556 5557\begin{example}[The cup product on a torus]5558Let \(X=T^2\) and take \(L\) to be the connected sum of two copies of5559\(S^1\times S^2\). The first cohomology groups of \(X\) and \(L\) both5560have rank two over \(\mathbb Z\). The product of the two coordinate5561classes in \(H^1(T^2;\mathbb Z)\) generates \(H^2(T^2;\mathbb Z)\).5562On \(L\), all products of degree-one classes are zero.5563Indeed, collapsing the separating sphere of the connected sum gives5564a map into the wedge of the two closed summands, inducing an5565isomorphism in \(H^1\) by the Mayer--Vietoris sequence.5566In the wedge, classes on different summands have zero product.5567Each summand has one degree-one generator pulled back from \(S^1\),5568so its square vanishes too. Thus the ring retraction displayed above5569cannot exist, even though comparing \(H^1\) alone would permit it.5570\end{example}5571 5572\section{A sharp uniform obstruction on the sphere}5573 5574\begin{proposition}[Reconstruction with a Euclidean intermediary]5575\label{prop:sphere-reconstruction}5576Let \(n>0\), and view \(S^n\) as the unit sphere in5577\(\mathbb R^{n+1}\). The following statements use the Euclidean5578norm on that ambient space.5579\begin{enumerate}5580\item For every \(m\), if \(e:S^n\to\mathbb R^m\) and5581  \(d:\mathbb R^m\to S^n\) are continuous, then5582  \(\sup_{x\in S^n}\|d(e(x))-x\|=2\).5583\item If \(d\) is allowed to land in \(\mathbb R^{n+1}\) and5584  \(m\le n\), the minimum possible uniform reconstruction error is \(1\).5585  For \(m\ge n+1\), there is an exact ambient reconstruction.5586\end{enumerate}5587\end{proposition}5588 5589\begin{proof}5590In the first case \(d\circ e\) has degree zero by contractibility.5591If it had no antipodal coincidence \(d(e(x))=-x\), then the straight5592segment between \(x\) and \(d(e(x))\), normalized to the unit sphere,5593would be a homotopy to the identity. Such a segment passes through5594the origin only for antipodal endpoints. The degrees therefore5595force an antipodal coincidence and the claimed error.5596 5597For the second part, the Borsuk--Ulam theorem applied to \(e\),5598after including \(\mathbb R^m\) into \(\mathbb R^n\), gives a pair5599\(x,-x\) with the same encoded value \cite{matousek-borsuk}.5600Both reconstructions are a common vector \(u\). The triangle5601inequality gives \(2\le\|u-x\|+\|u+x\|\), so at least one error5602is at least \(1\). The constant reconstruction \(0\) attains5603that bound. For \(m\ge n+1\), include the sphere in the first5604\(n+1\) coordinates and decode by their projection.5605\end{proof}5606 5607The distinction between the codomains matters. If a reconstruction5608lands close to the sphere, one can normalize along its realized5609image. This does not imply that the decoder can be normalized on5610the entire intermediate \(\mathbb R^m\), where it may take value zero.5611 5612\section{Spaces of factorizations}5613 5614In the category of spaces the factorizations of \(f\) through \(L\),5615together with their chosen homotopies, form the homotopy pullback5616\[5617\begin{tikzcd}[column sep=large]5618\operatorname{Fact}_L(f) \arrow[r] \arrow[d] &5619\operatorname{Map}(X,L)\times\operatorname{Map}(L,Y)5620  \arrow[d,"\circ"]\\5621\{f\} \arrow[r,hook] & \operatorname{Map}(X,Y) .5622\end{tikzcd}5623\]5624A point is a triple \((e,d,\gamma)\), with \(\gamma\) a path from5625\(d\circ e\) to \(f\). Paths between such points track changes5626of both maps and homotopies. This is a homotopy problem, while an5627exact smooth factorization requires equality \(d\circ e=f\) and5628smooth maps. For example, the smooth map5629\(e^{it}\mapsto \exp(i\sin t)\) on \(S^1\) is nullhomotopic and5630therefore factors through a point in the homotopy sense. It is5631nonconstant, so it does not admit an exact factorization through a5632point.5633 5634Families can have an obstruction even when every individual target5635value has a lift. With decoder \(d(z)=z^2\) on \(L=S^1\), every5636single \(b\in S^1\) has a preimage. The family \(f(b)=b\) has no5637continuous choice of encoded value \(e(b)\) satisfying5638\(d(e(b))=b\). Taking degrees would yield \(2\deg(e)=1\), which