We consider the notion of a tangle and analyze the operations which are possible on it. We use tangles in constructing a new notation for knots based on Conway’s knot notation. This new notation has several advantages over existing notations. All the basic properties of the notation and algorithm s to retrieve simple knot inform ation are discussed. P rocedures for puttin g a knot into our notation are also given. Finally, polynomial-tim e algorithm s, which do not rely on topological deform ation, are described which produce a plait and a closed braid which are isotopic to any knot given in our notation.
F irst, we review th e notion of tangles and investigate th eir classification. We shall then intro duce the new notation, prove th a t all knots may be represented by it, give an algorithm to place a given knot into this notation and present a traversal algorithm which will calculate certain features of the knot. An algorithm is th en given to obtain a plait and a braid, th e closures of which are a given knot in the new notation.
5.1
Tangles
5.1.1
D efin ition and P a r titio n
Consider the 3-ball and choose 2n points on its surface, which is th e 2-sphere 5^, and call the set of these points P . A ttach n polygonal curves to the 2n points such th a t: (i) each curve intersects 5^ in exactly 2 points in P , which are its endpoints, (ii) exactly one curve m ay begin or end at any one point in P and (iii) no curve m ay intersect another. If th e set of these curves is T , then we will call th e set an n-tangle. In particular, we will focus on 2-tangles and so whenever we skip the n, it will be understood th a t we m ean n = 2. N ote th a t our requirem ent th a t the curves be polygonal excludes any wild tangles, where wild is to be understood in the usual knot theory sense. Two tangles are called equal if they are isotopic w ithout moving th e points in P .
5.1 T angles 93
N E N W
S E
sw
Figure 5.1: The 3-ball and th e four points on its surface which form the endpoints of the two polygonal curves necessary to define a tangle.
points of the compass)
W = ( 0 , - i
^ (5.1)(5.2)
on the unit sphere, which will be our canonical see figure 5.1. Even though tangles are, by definition, three dimensional objects, we will work w ith their projection onto th e two dimensional plane as if the projection is th e tangle. The fact th a t a projection in which there are a t worst double points always exists for a tangle follows from the corresponding theorem about knots.
Figure 5.2: The elem entary tangles.
We shall find it convenient to partitio n the set of all possible tangles into a few categories: elementary, integral, fractional, rational and irrational. T he sim plest are the elementary tangles, of which there are four. These are best introduced by displaying them in figure 5.2. Note th a t we have not drawn B ^, it should however be understood to be present. The reason for nam ing them as they have been will become apparent later on. Note th a t th e literatu re disagrees on which of the two tangles ±1 is to have the minus sign, this is a m atter of convention and has no serious consequences (we follow the convention introduced by Conway).
The other types of tangles can be m ost readily defined in term s of combining the elem entary ones in some way. To do this, we shall define two ways of adding tangles. Following Conway, we denote a general tangle by an ”L” shaped symbol within th e 3-ball and we also sketch the ends of the two curves by which tangles m ay be attached to one another. In this way, we define the horizontal sum -f and the vertical sum © in figure 5.3.
In w hat follows, we shall use a superscript to denote of which type a particular tangle t is; for
K n o ta tio n an d B ra id in g a K n o t 94
B B
Figure 5.3: Tangle addition.
example an elem entary tangle t would be denoted by An integral tangle and a fractional tangle t^^^ will be defined in term s of the elem entary tangles ± 1 by
= 1 + 1 + . . . + 1 (5.3)
t factors
t(/) = 1 0 1 0 . . . 0 1 (5.4)
'--- V--- ' t factors
The negative versions are, of course, the sums of —1 tangles instead of 1 tangles. A rational tangle can then be defined in term s of a sum of integral and fir actional tangles. The definition of the sum differs if the num ber of tangles j in th e sum is even or odd, this is because the definition requires an alternate sum between integral and fractional (and th e two m ethods of addition) which always ends in an integral tangle being added. This is because th e set of rational tangles m ay be classified if this restriction is imposed; the classification scheme is outlined in the next section. The integral tangles, including th e last, m ay be zero and the fractional tangles m ay be infinite. If any component tangles are 0 or oo though, they m ay be removed from the sum and the term s im mediately preceding and following the removed term m ay be added together to shorten the sum, while preserving isotopy.
f W = + c(') © + . . . + (5.5) ^ ^ / j odd tW = n ( ^ ) + 6 M © c ( ^ ) + d W © . . . + %(') (5.6) V---^--- / j even
Note th a t the set of elem entary tangles is a subset of both th e integral and fractional tangle sets which are subsets of th e rational tangle set. We shall call any tangle which is not rational, irrational.
5.1.2
C lassification o f Tangles
We m ay denote a rational tangle by giving its integral and fractional factors in order. T hus a sequence of integers = { a i,a2, . Ui) defines any rational tangle. Note again th a t the identity of the tangle factors is decided by requiring the last in the sequence to be integral. Given a rational tangle = (a i, û2, • • • , «i), we m ay associate w ith it an extended rational num ber E{t^'^^) = a / (3,
where a and (3 are integers including zero. We say an extended rational num ber because this allows for 1/0 = oo, the inclusion of which extends the rational numbers. We calculate E (((^)) by the
5.2 K n o t N o ta t io n 95
continued fraction (the + signs are arithm etic additions and not tangle additions)
E («(’■)) = aj + --- (5.7)
O ^ i - 1 H---
ai-2 • • • --- Oi
Conway [50] was able to deduce th a t two rational tangles are equal if and only if the associated extended rational num bers were equal, this is called Conway’s Basic Theorem. The first published proof may be found in [39] b u t a more intuitive proof was given by Goldm an and Kauffman [71]. Thus Conway’s Basic Theorem classifies rational tangles in a simple algorithm ic m anner.
In particular, the fractions associated w ith th e elem entary tangles are their num erical names: 0, ±1 and oo. The fraction for an integral tangle is and for a fractional tangle is It is clear now why these tangles were nam ed as they were. T his concludes our review of previous work on tangles and the rest of th e chapter is new work.
By equation 5.7 is easy to calculate the fraction associated w ith a given rational tangle. Given a fraction, it is also possible to decompose it into appropriate factors, thereby constructing the rational tangle associated w ith it. Euclid’s algorithm will accomplish this.
5.2
K not N otation
Tangles were invented in an effort classify knots (they m ay be used to classify two-bridge knots via the correspondence w ith the extended rational num bers [116]) and so we m ust have a m ethod to combine tangles into knots. Conway [50] showed th a t any knot m ay be obtained by substituting several rational tangles into the vertices of basic polyhedra. A polyhedron, in the sense of Conway, is an edge-connected 4-valent planar m ap and it is basic if, in addition, no region (including the infinite region) has ju st two vertices. Conway constructs the 8 different basic polyhedra necessary to denote all prime knots up to and including 11 crossings. T he b eauty of using the basic polyhedra is th a t small knots m ay be nam ed quite efficiently, th a t is one gives the basic polyhedron and the tangle fractions to be substituted. However it can be quite a chore to construct the Conway nam e of a large knot. The next section will introduce our new knot notation.
5.2.1
T he U n iversal P olyh ed ron
Consider the basic polyhedron P { i ,j ) shown in figure 5.4; we will call it the universal polyhedron. It is a prototype for a knot projection. The circles will be called vertices and the lines connecting them edges. The vertices are arranged into i rows of j vertices each. Each vertex can thus be labeled by its row and column index. While P { i ,j ) denotes th e whole polyhedron and specifies the number of rows and columns, pki specifies a particular vertex in row k and column I. In w hat follows, we will substitute rational tangles into th e vertices to yield a knot projection. Since a rational tangle m ay be specified by a single extended rational num ber, pki takes an extended rational number value. By substituting rational tangles into all vertices of a given polyhedron, we
K n o ta tio n an d B ra id in g a K n o t 96
• • •
• • •
• • •
Figure 5.4; The universal polyhedron
obtain a knot projection of some tam e knot. This can be completely specified by giving all pki a value, which m ay be arranged into a m atrix form,
P i i P u