cw-functions-tutorial.xml

<Chapter><Heading>Regular CW-structure on knots</Heading>

        <Section><Heading>Knot complements in the 3-ball</Heading>

                While methods for endowing knot complements with CW-structure already exist in HAP (see section 2.1), they often result in a large
                number of cells which can make computing with them taxing. The following example shows how one can obtain a comparatively small 3-dimensional regular CW-complex
                corresponding to the complement of a thickened trefoil knot from an arc presentation. Recall that an arc presentation is encoded in HAP as
                a list of integer pairs corresponding to the position of the endpoints of each horizontal arc in a grid.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-01.txt">
</Example>

                An optional argument of <C>"rand"</C> in the <C>KnotComplement</C> function randomises the order in which <M>2</M>-cells are added to the complex.
                This allows for alternate presentations of the knot group.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-02.txt">
</Example>

                It is often useful to obtain an inclusion of regular CW-complexes <M>\iota : \partial (N(K)) \hookrightarrow B^3 \backslash N(K)</M> from the boundary of a
                tubular neighbourhood of some knot <M>N(K)</M> into its complement in the <M>3</M>-ball <M>B^3 \backslash N(K)</M>. The below example does this for the first
                prime knot on 11 crossings.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-03.txt">
</Example>

                Note that we can add <M>n</M>-cells to regular CW-complexes by specifying the <M>(n-1)</M>-cells in their boundaries and <M>(n+1)</M>-cells in their coboundaries.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-10.txt">
</Example>

        </Section>
        <Section><Heading>Tubular neighbourhoods</Heading>

                Let <M>Y</M> denote a CW-subcomplex of a regular CW-complex <M>X</M> and let <M>N(Y)</M> denote an open tubular neighbourhood of <M>Y</M>.
                Given an inclusion of regular CW-complexes <M>f : Y \hookrightarrow X</M>, this algorithm describes a procedure for
                obtaining the associated inclusion <M>f' : \partial C \hookrightarrow C</M> where <M>C=X \backslash N(Y)</M> and <M>\partial C</M> denotes
                the boundary of <M>C</M>. The following is also assumed:
                <P/>Let <M>e^n</M> denote a cell of <M>X \backslash Y</M> of dimension <M>n</M> with <M>\bar{e}^n</M> denoting its closure. For each <M>n</M>-cell, all of
                the connected components of the subcomplex <M>\bar{e}^n \cap Y</M> are contractible.
                <P/>Some additional terminology and notation is needed to describe this algorithm. The output regular CW-complex <M>X \backslash N(Y)</M> consists of the
                cell complex <M>X \backslash Y</M> as well as some additional cells to maintain regularity. A cell of <M> X \backslash N(Y)</M> is referred to as
                <E>internal</E> if it lies in <M>X \backslash Y</M>, it is <E>external</E> otherwise. Let <M>\bar{e}^n</M> denote the closure in <M>X</M> of an internal cell <M>e^n</M>.
                Note that <M>\bar{e}^n</M> is a CW-subcomplex of <M>X</M> and so is the intersection <M>\bar{e}^n \cap Y</M> which can be
                expressed as the union
                <P/><M>\bar{e}^n \cap Y = A_1 \cup A_2 \cup \cdots \cup A_k</M>
                <P/> of its path components <M>A_i</M> all of which are CW-subcomplexes of <M>Y</M>. For each <M>n</M>-cell of <M>X \backslash Y</M> there is one
                internal <M>n</M>-cell <M>e^n</M> of <M>X \backslash N(Y)</M>. For <M>n \geq 1</M> there is also one external <M>(n-1)</M>-cell <M>f^{e^n}_{A_i}</M> for each path
                component <M>A_i</M> of <M>\bar{e}^n \cap Y</M>. Lastly, we need a method for determining the homological boundary of the internal and external
                cells:
                <P/><M>\bullet</M> The boundary of an internal <M>n</M>-cell <M>e^n</M> consists of all those internal <M>(n-1)</M>-cells of <M>\bar{e}^n</M>
                together with all external <M>(n-1)</M>-cells <M>f^{e^n}_{A_i}</M> where <M>A_i</M> is a path component of <M>\bar{e}^n \cap Y</M>.
                <P/><M>\bullet</M> The boundary of an external <M>(n-1)</M>-cell <M>f^{e^n}_{A_i}</M> consists of all those external <M>(n-2)</M>-cells
                <M>f^{e^{n-1}}_{B_j}</M> where <M>e^{n-1}</M> is an <M>(n-1)</M>-cell of <M>\bar{e}^n</M> and <M>B_j \subseteq A_i</M> is a path component of
                <M>A_i</M>.
                <P/>The following three steps comprise the algorithm.
                <P/><M>(1)</M> For each internal <M>n</M>-cell <M>e^n \subset X \backslash Y</M>, compute the CW-complex <M>\bar{e}^n \cap Y</M> as a union of
                path components <M>A_1 \cup A_2 \cup \cdots \cup A_k</M>. This information can be used to determine the number of cells of <M>X \backslash N(Y)</M>
                in each dimension.
                <P/><M>(2)</M> Create a list <M>B=[ \; [ \; \; ], [ \; \; ], \ldots, [ \; \; ] \; ]</M> of length <M>\textrm{dim}X +1</M>.
                <P/><M>(3)</M> For <M>0 \leq n \leq \textrm{dim}X</M> set <M>B[n+1]=[ b_1, b_2, \ldots, b_{\alpha_n} ]</M> where <M>\alpha_n</M> is the number
                of <M>n</M>-cells in <M>X \backslash N(Y)</M> and <M>b_i</M> is a list of integers describing the <M>(n-1)</M>-cells of the <M>i ^ \textrm{th}</M>
                <M>n</M>-cell of <M>X \backslash N(Y)</M>. The internal cells will always be listed before the external cells in each sublist. Return B as a regular CW-complex.
                <P/> The following example computes the tubular neighbourhood of a <M>1</M>-dimensional subcomplex of a <M>3</M>-dimensional complex corresponding to the Hopf link
                embedded in the closed <M>3</M>-ball.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-04.txt">
</Example>

                Note that the output of this algorithm is just a regular CW-complex, not an inclusion map. The function <C>BoundaryMap</C> can be employed to obtain the
                boundary of a pure complex. This results in three path components for this example: two corresponding to the boundary of the knotted tori and the
                other corresponding to the boundary of the <M>3</M>-ball in which the link was embedded. These path components can be obtained as individual CW-subcomplexes if desired.
                A CW-subcomplex is represented in HAP as a list <M>[X,s]</M> where <M>X</M> is a regular CW-complex and <M>s</M> is a list of length <M>n</M> whose
                <M>i^\textrm{th}</M> entry lists the indexing of each <M>(i-1)</M>-cell of the <M>n</M>-dimensional subcomplex of <M>X</M>. CW-subcomplexes and CW maps can be converted between each other interchangeably.
                This next example obtains the inclusion detailed in the above algorithm, finds the path components of the source of said inclusion, shows that they are in fact disjoint, and then
                obtains the first four integral homology groups of each component.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-05.txt">
</Example>

                As previously mentioned, for the tubular neighbourhood algorithm to work, we require that no external cells yield non-contractible path-components in their
                intersection with the subcomplex. If this is ever the case then we can subdivide the offending cell to prevent this from happening. We have implemented two
                subdivision algorithms in HAP, one for barycentrically subdividing a given cell, and the other for subdividing an <M>n</M>-cell into as many <M>n</M>-cells as 
                there are <M>(n-1)</M>-cells in its boundary. Barycentric subdivision is integrated into the <C>RegularCWComplexComplement</C> function and will be performed automatically as
                required. The following example shows this automatic subdivision running via the complement of a tubular neighbourhood of the unknot, then obtains an inclusion map from the closure
                of an arbitrary <M>3</M>-cell of this complex and then compares the difference in size of the two different subdivisions of a 2-cell in the boundary of this <M>3</M>-cell.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-06.txt">
</Example>

        </Section>
        <Section><Heading>Knotted surface complements in the 4-ball</Heading>

                A construction of Satoh's, the tube map, associates a ribbon torus-knot to virtual knot diagrams. A virtual knot diagram differs from a knot diagram in that it allows for a third
                type of crossing, a virtual crossing. The image of such a crossing via the tube map is two tori which pass through each other. An arc diagram is a triple of lists <C>[arc,cross,cols]</C>
                that encode virtual knot diagrams. <C>arc</C> is an arc presentation. <C>cross</C> is a list of length the number of crossings in the knot associated to the arc presentation whose entries
                are <M>-1,0</M> or <M>1</M> corresponding to an undercrossing (horizontal arc underneath vertical arc), a virtual crossing (depicted by intersecting horizontal and vertical arcs) and an overcrossing
                (horizontal arc above vertical arc) respectively. <C>cols</C> is a list of length the number of <M>0</M> entries in <C>cross</C> and its entries are <M>1,2,3</M> or <M>4</M>.
                It describes the types of 'colourings' we assign to the virtual crossings. We interpret each integer as the change in 4-dimensional height information as represented by a colour
                scale from blue (lower down in 4-space), to green (0 level), to red (higher up in 4-space). Without loss of generality, we impose that at each virtual crossing, the vertical arc passes through the horizontal arc.
                Thus, <M>1</M> corresponds to the vertical bar entering the horizontal bar as blue and leaving as blue, <M>2</M> corresponds to entering as blue and leaving as red, <M>3</M>
                corresponds to entering as red and leaving as blue and <M>4</M> corresponds to entering and leaving as red. A coloured arc diagram can be visualised using the <C>ViewColouredArcDiagram</C>
                function.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-07.txt">
</Example>
<P/>
<Alt Only="HTML">&lt;img src="../lib/Kelvin/images/kelvin-01.png" width="350"/></Alt>
<P/>

                Towards obtaining a regular CW-decomposition of ribbon torus-knots, we first begin by embedding a self-intersecting knotted torus in the 3-ball. The function 
                <C>ArcDiagramToTubularSurface</C> inputs a coloured arc diagram and outputs an inclusion from the boundary of some (potentially self-intersecting) torus in the <M>3</M>-ball.
                By inputting just an arc presentation, one can obtain an inclusion identical to the <C>KnotComplementWithBoundary</C> function. By additionally inputting a list of <M>-1</M>s
                and <M>1</M>s, one can obtain an inclusion similar to <C>KnotComplementWithBoundary</C> but where there is extra freedom in determining whether or not a given crossing
                is an under/overcrossing. If one inputs both of the above but includes <M>0</M> entries in the <C>cross</C> list and includes the list of colours, the output is then an
                inclusion from an embedded self-intersecting torus into the 3-ball where each <M>2</M>-cell (the top-dimensional cells of the self-intersecting surface) is assigned a colour.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-08.txt">
</Example>
                From this self-intersecting surface with colour, we can lift it to a surface without self-intersections in <M>\mathbb{R}^4</M>. We do this by constructing a regular CW-complex
                of the direct product <M>B^3 \times [a,b]</M> where <M>B^3</M> denotes the <M>3</M>-ball, <M>a</M> is <M>1</M> less than the smallest integer assigned to a cell by the colouring, and
                <M>b</M> is <M>1</M> greater than the largest integer assigned to a cell by the colouring. The subcomplex of the direct product corresponding to the surface without intersection 
                can be obtained using the colouring with additional care taken to not lift any 1-cells arising as double-point singularities. The following example constructs the complement of a
                ribbon torus-link embedded in <M>\mathbb{R}^4</M> obtained from the Hopf link with one virtual crossing and then calculates some invariants of the resulting space. We compare the
                size of this complex, as well as how long it takes to obtain the same invariants, with a cubical complex of the same space. As barycentric subdivision can massively increase the size of the cell complex,
                the below method sequentially obtains the tubular neighbourhood of the entire subcomplex by obtaining the tubular neighbourhood of each individual <M>2</M>-cell. This has yet to be optimised so it currently
                takes some time to complete.

<Example>
<#Include SYSTEM "../lib/Kelvin/tutex/kelvin-09.txt">
</Example>
                
        </Section>       
</Chapter>