Characteristic class of isotopy for surfaces (Q2661857): Difference between revisions

The paper studies piecewise-smooth (PS) isotopy invariants of PS embeddings of 3-dimensional surfaces into the 3-dimensional Euclidean space \(\mathbb R^3\). For a closed oriented surface \(M\) there are two natural invariants. For simplicity I define them for smooth embeddings \(f:M\to\mathbb R^3\). Take a shift \(\overline f:M\to\mathbb R^3\) so that for any \(x\in M\) the vector \(\overline f(x)-f(x)\) is a small length vector normal to \(f(M)\) agreeing with the orientation of \(M\) and of \(\mathbb R^3\). Let \(I(f)\) be \(+1\) if this vector looks outside \(f(M)\), and \(-1\) otherwise. Take any two integer 1-dimensional cycles (e.g. oriented closed curves) \(b,c\) on \(M\). Let \(L(f)(b,c)\) be the linking number of \(f|_b\) and \(\overline f|_c\). This defines a bilinear form \(L(f):H_1(M)\times H_1(M)\to\mathbb Z\) called Seifert form. (We omit \(\mathbb Z\)-coefficients of homology groups.) The form is \(\cap_M\)-symmetric, i.e., \(L(f)(\beta,\gamma)-L(f)(\gamma,\beta)=\beta\cap_M\gamma\), where \(\cap_M\) is the algebraic intersection of homology classes on \(M\). Since \(M\) is homeomorphic to the sphere with \(g\) handles, \(H_1(M)\) has a basis of \(2g\) elements. The matrix of \(L(f)\) in this basis consists of \(4g^2\) integers and is skew-symmetric. So the invariant \(L(f)\) amounts to a collection of \(g(2g+1)\) integer invariants. Presumably these integers are independent, i.e., any \(\cap_M\)-symmetric bilinear form \(H_1(M)\times H_1(M)\to\mathbb Z\) is realizable as \(L(f)\) for some \(f\). These two invariants can be derived from the more general Haefliger-Wu invariant \(\alpha(f)\), see e.g. the survey [\textit{A. B. Skopenkov}, Lond. Math. Soc. Lect. Note Ser. 347, 248--342 (2008; Zbl 1154.57019), \S5]. I conjecture that the Haefliger-Wu invariant amounts to the above two invariants, i.e., \(\alpha(f)=\alpha(f')\) if and only if \(I(f)=I(f')\) and \(L(f)=L(f')\). I have to warn the reader that the paper under review describes a material similar to the above in a confusing and even erroneous way. Here are the most important critical remarks. In \S6 instead of embeddings, only their images are considered, which are not sufficient for calculating the Haefliger-Wu invariant. Instead of giving a \(\{+1,-1\}\)-valued invariant and four (dependent) integer-valued invariants for the torus, \S6 gives five real-valued invariants. The mistake is just before Algorithm 4 in p. 2150: `the volume of an element of \(H^2(S^2)\)' is meaningless. The definition would make sense if one defines \(V(\widetilde f(h_i))\) to be the integer \(k\) such that \((\widetilde f)^*(h_i)=k\omega\), where \(\omega\) is the standard generator of \(H^2(S^2)\). As opposed to the beginning of \S5, the collection of the above invariants (and the Haefliger-Wu invariant) are not complete for isotopy, see e.g. the survey [\textit{A. Skopenkov}, \url{http://www.map.mpim-bonn.mpg.de/Embeddings_in_Euclidean_space:_an_introduction_to_their_classification}, Remark 6.3.b]. In \S5 the Haefliger-Wu invariant \(\alpha(f)\) is without explanation replaced by the potentially weaker (although possibly equivalent) invariant \((\widetilde f)^*(\omega)\) (observe that the expression \(\widetilde{f^*\ }(\omega)\) of the paper is meaningless). The references in the last paragraph of \S2 are misleading. The paper [\textit{D. Gonçalves} and \textit{A. Skopenkov}, Topology Appl. 153, No. 12, 2026--2034 (2006; Zbl 1105.57022)] is on the existence not on the classification problem. The paper [\textit{D. Tonkonog}, ibid. 158, No. 9, 1157--1162 (2011; Zbl 1235.57003)] does not concern the Haefliger-Wu invariant; perhaps [loc. cit.] was confused with [\textit{D. Tonkonog}, ``Embedding punctured \(n\)-manifolds in Euclidean \((2n-1)\)-space'', Preprint, \url{arXiv:1010.4271}]. No reference to the above well-known definitions of the \(\{+1,-1\}\)-valued invariant and of the Seifert form are presented.