On the combinatorics of smoothing (Q289736)

The focus of this article is loop counting and its applications. Many knot invariants are computed by smoothing the crossings of a knot diagram and counting the resulting components. The Refined Loop Counting Principle (RLCP) counts the number of components in a partially resolved knot diagram. Given an oriented virtual knot diagram \(K\), the crossings of the diagram are divided into three subsets. The subset \(S_O\) denotes the crossings that are smoothed following the orientation of the diagram, the subset \(S_U\) denotes the set of unoriented smoothings, and \(S_\emptyset \) denotes the set of crossings that are not smoothed. The triple \(S=(S_O, S_U, S_{ \emptyset} )\) is called a partial smoothing of the virtual diagram. In the Gauss diagram \(D\) corresponding to \(K\), remove all arrows corresponding to elements of \(S_{ \emptyset} \) to form the Gauss diagram \(D_{ \emptyset} \). For a square matrix \(A\), \(m_0 (A)\) denotes the multiplicity of zero as a root of the characteristic polynomial of \(A\). For a based Gauss diagram \(D\), there is a corresponding linearly ordered intersection graph \(\vec{G}\). From this graph, the skew-adjacency matrix of \(\vec{A}_{\vec{G}} \) is obtained. The notation \( \sharp_D (S)\) denotes the number of curves in the partial state \(S\). The RLCP is stated as: Let \(S=(S_O, S_U, S_{ \emptyset} )\) be a partial state and let \( \vec{G}_{ \emptyset} \) denote the linearly ordered graph of \(D_{ \emptyset} \). The matrix \(\vec{A} _{ \vec{G_{\emptyset}}}\) is the matrix obtained from \( \vec{G}_{\emptyset} \). Then: 1) if \(S_U= \emptyset \) then \( \sharp_D (S) = m_0 ( \vec{A}_{\vec{G_{\emptyset}}} ) + 1 \), 2) if \( j \in S_U \neq \emptyset\), then there are Gauss diagrams \(D_j ^f (S) \) and \( D^s _j (S) \), called double covers of \(D\), and these double covers have the property that \( \sharp_{D_j^f (S)} (O) = 2 \sharp_D (S) = \sharp_{D_j^s (S)} (O)\) and 3) \(m_0 ( \vec{G} _{\vec{A}} )\) may be computed directly from simpler linearly ordered graphs. This theorem is a refinement of Zulli's loop counting principle [\textit{L. Traldi} and \textit{L. Zulli}, Topology Appl. 156, No. 15, 2515--2526 (2009; Zbl 1179.57005)]. The RLCP reduces the counting problem to finding the characteristic polynomial of the skew-adjacency matrix of the linearly ordered graph associated with a Gauss diagram. In the cases when \(S_u\) is non-oriented, a double cover of the Gauss diagram is used. A substantial portion of the paper focuses on extending results from spectral graph theory to linearly ordered graphs. The RLCP is then applied to answer the following question: Given a virtual knot diagram, how many ways are there to give an oriented smoothing on \(m-j\) crossings and an unoriented smoothing on \(j\) crossings so that the result has exactly one connected component? For fixed values of \(j\) and \(m\), this is called Question \((j,m)\). The questions \((0,m)\) and \((1,m)\) are solved for pretzel knots \(L(p,q,r) \) with all \(p,q,\) and \(r\) odd and the case where one of \(p,Q\), and \(r\) is even.