Independent scholarship · Open accessThe monthly digest

Indexed text attachment

algebraic-and-differential-topology-in-higher-categories.tex

Source for Algebraic and Differential Topology in Higher Categories

View paper · Download original file

Lines 9301–9400 of 10051. Source text is displayed literally.

9301\ifnum\qa=1 \draw[very thick,bwd] (0,1.1)--(2.2,1.1); \else \draw[thick,densely dashed,ink2] (0,1.1)--(2.2,1.1); \fi9302\ifnum\qb=1 \draw[very thick,bwd] (1.1,0)--(1.1,2.2); \else \draw[thick,densely dashed,ink2] (1.1,0)--(1.1,2.2); \fi9303\ifnum\qc=1 \draw[very thick,bwd] (0,0)--(2.2,2.2); \else \draw[thick,densely dashed,ink2] (0,0)--(2.2,2.2); \fi9304\node at (1.1,-0.4) {\small$\Arf=\arf$};9305\end{scope}}9306\end{tikzpicture}9307\figcap{The four quadratic refinements on a torus, drawn as a square with opposite sides identified. The curves9308$a$ (horizontal), $b$ (vertical) and $a+b$ (diagonal) are solid where $q=1$ and dashed where $q=0$. Three forms9309have one solid curve; they are permuted by the symmetries of the torus. The fourth has all three, and nothing9310can be cut.}\label{fig:forms}9311\end{center}9312 9313\section{Surgery}9314 9315\begin{center}9316\begin{tikzpicture}[scale=0.92]9317\begin{scope}9318\draw[thick,fill=fwd!10] (0,0) ellipse (1.9 and 1.05);9319\draw[thick] (-0.75,0.08) arc (200:340:0.8 and 0.4);9320\draw[thick] (-0.55,-0.05) arc (160:20:0.59 and 0.26);9321\draw[very thick,bwd] (0,-0.27) arc (90:270:0.14 and 0.39);9322\draw[very thick,bwd,densely dashed] (0,-0.27) arc (90:-90:0.14 and 0.39);9323\draw[very thick,accent] (0,0.05) ellipse (1.3 and 0.62);9324\node[bwd] at (0.4,-0.75) {\small$a$}; \node[accent] at (1.45,0.55) {\small$b$};9325\node at (0,-1.5) {\small torus, $q(a)=0$};9326\end{scope}9327\draw[->,thick,ink2] (2.3,0)--(3.1,0);9328\begin{scope}[xshift=5.3cm]9329\draw[thick,fill=fwd!10] (-1.9,0) arc (180:20:1.9 and 1.05) -- (1.78,-0.36) arc (-20:-180:1.9 and 1.05);9330\draw[thick,fill=white] (1.78,0.36) arc (90:270:0.14 and 0.36);9331\draw[thick] (1.78,0.36) arc (90:-90:0.14 and 0.36);9332\draw[thick] (-0.75,0.08) arc (200:340:0.8 and 0.4);9333\draw[thick] (-0.55,-0.05) arc (160:20:0.59 and 0.26);9334\node at (0,-1.5) {\small cut along $a$};9335\end{scope}9336\draw[->,thick,ink2] (7.6,0)--(8.4,0);9337\begin{scope}[xshift=10.3cm]9338\draw[thick,fill=fwd!10] (0,0) ellipse (1.5 and 1.0);9339\draw[thick,densely dashed] (-1.5,0) arc (180:0:1.5 and 0.3);9340\draw[thick] (-1.5,0) arc (180:360:1.5 and 0.3);9341\node at (0,-1.5) {\small cap with two discs};9342\end{scope}9343\end{tikzpicture}9344\figcap{Surgery, which the source calls reconstruction. A curve $a$ on the surface with $q(a)=0$ carries the frame9345that extends over a disc. Cut along $a$ and glue in two parallel copies of that disc, with the extended frame:9346the result is a framed surface of genus one less, framed cobordant to the original. If $q(a)=1$ the frame does9347not extend and the handle cannot be cut.}\label{fig:surgery}9348\end{center}9349 9350\begin{proposition}\label{prop:surgery}9351\Scl[\src{II,\,\S11, \S15}] Surgery on a class $a$ with $q(a)=0$ replaces $H_1$ by $a^\perp/\langle a\rangle$ with the9352induced form, and does not change the Arf invariant. A form with $\Arf=0$ can be reduced to genus zero. A form9353with $\Arf=1$ can be reduced to a torus on which $q$ is $1$ on all three non-zero classes, and no further.9354\end{proposition}9355 9356\noindent This was verified for all $84$ forms of genus at most three, by carrying out the reduction. \chk{A5}9357 9358\begin{theorem}\label{thm:second}9359\Scl[\src{II,\,\S15, Thm.~24}] For $n\ge2$, $\pi_{n+2}(S^n)\cong\Z/2$. The class of a framed surface is the Arf9360invariant of its quadratic form, and the non-zero class is represented by the torus $S^1\times S^1$ with the9361product of two copies of the non-bounding frame of the circle: $\eta^2\ne0$.9362\end{theorem}9363\begin{proof}[Structure of the proof]9364By \Cref{prop:surgery} a framed surface is framed cobordant to a sphere or to the odd torus. A framed sphere9365is framed cobordant to zero, since $\pi_2(\SO(n))=0$ and so all frames on it are equivalent. The Arf invariant is9366a cobordism invariant, by the analysis of the critical levels of a framed $3$-manifold \src{II,\,\S15,9367Thms.~22--24}, so the odd torus is not zero.9368\end{proof}9369 9370\section{History, and what the form means}9371 9372The first announcement of this result, in 1938, gave the answer zero; the correct value was published in 19509373\src{II, refs.~1, 2}. The slip was to treat $q$ as if it were linear. If it were, it would vanish on a9374half-dimensional subspace, every handle could be cut, and the group would be trivial. The cross term $x\cdot y$9375in \Cref{thm:quadratic} is what prevents this, and it has a structural meaning.9376 9377\begin{remark}[Why quadratic]\label{rem:kinvariant}9378\Scl A class $x\in H_1(M;\Z/2)$ on a framed surface determines a framed circle, hence an element $q(x)$ of9379$\pis_1$. So $q$ is the map on first framed bordism induced by collapsing $M$ to a point, restricted to classes9380coming from homology, and it would be linear if the framed bordism of $M$ split as the homology of $M$ with9381coefficients in $\pis_0$ and $\pis_1$ separately. It does not split: the sphere spectrum is not a product of9382Eilenberg--MacLane spectra, and the first obstruction, its first $k$-invariant, is the Steenrod square $Sq^2$.9383The Arf invariant of a framed surface is that $k$-invariant, seen through a manifold.9384\end{remark}9385 9386\begin{remark}[Spin structures]\label{rem:spin}9387\Sli[\cite{Joh80,Ati71}] A stable framing of a surface induces a spin structure, and $q$ depends only on it: the9388quadratic refinements of the intersection form correspond exactly to the $2^{2g}$ spin structures on the9389surface. The Arf invariant is the parity of the dimension of the space of holomorphic sections of the9390corresponding square root of the canonical bundle, and is the isomorphism from the spin bordism group of9391surfaces to $\Z/2$. The counts of \Cref{thm:arf}\emph{(b)} are the numbers of even and odd theta9392characteristics.9393\end{remark}9394 9395\begin{remark}[The Kervaire invariant]\label{rem:kervaire}9396\Sli[\cite{Ker60,Bro69,HHR16,LWX24}] The same construction on framed manifolds of dimension $4m+2$, with the middle9397homology in place of $H_1$, gives the Kervaire invariant. Browder showed that it can be non-zero only in9398dimensions $2^{j+1}-2$. Framed manifolds with invariant one exist in dimensions $2,6,14,30,62$; Hill, Hopkins and9399Ravenel showed there are none in dimensions $254$ and above. The preprint of Lin, Wang and Xu establishes9400existence in dimension $126$, the final case left by those results. The dimension-two case is \Cref{thm:second}.