See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222476293

Rational Tangles

Article in Advances in Applied Mathematics · April 1997

DOI: 10.1006/aama.1996.0511

CITATIONS READS

70 567

2 authors, including:

Louis Kauffman
University of Illinois Chicago
684 PUBLICATIONS 14,281 CITATIONS

SEE PROFILE

All content following this page was uploaded by Louis Kauffman on 09 January 2019.

The user has requested enhancement of the downloaded file.

ADVANCES IN APPLIED MATHEMATICS 18, 300]332 Ž1997.
ARTICLE NO. AM960511

Rational Tangles
Jay R. Goldman

School of Mathematics, Vincent Hall, Uni¨ ersity of Minnesota, Minneapolis,

Minnesota 55455

and

Louis H. Kauffman

Department of Mathematics, Uni¨ ersity of Illinois at Chicago, 851 South Morgan Street,
Chicago, Illinois 60607-7045

Received August 25, 1996

This paper gives an elementary and self-contained proof of Conway’s Basic

Theorem on rational tangles. This theorem states that two rational tangles are
topologically equivalent if and only if they have the same associated rational
fraction. Our proof divides into a geometric half that relates the arithmetic of
continued fractions to the topology of tangles and an algebraic part that defines
the fraction of any tangle via the bracket model for the Jones polynomial. We
present an application to molecular biology. Q 1997 Academic Press

1. INTRODUCTION

We give an elementary and self-contained proof of J. H. Conway’s Basic

Theorem on Rational Tangles. Conway associated a finite continued
fraction to each rational tangle. The sum of this continued fraction is the
fraction of the tangle. Conway’s Theorem states that two rational tangles
are ambient isotopic if and only if their fractions are equal. The meanings of
these terms will be explained in the body of the paper. Conway’s Theorem
is first stated Žwithout proof. in his paper w3x. Proofs of the theorem using a
fair amount of mathematical machinery have appeared in the literature of
knot theory Žsee w2x.. In the meantime, the subject of tangles has become
of wider interest due to its applications to the topology of DNA. We are
pleased with the elementary nature of our proof and we believe that it
provides a good place to begin learning these aspects of knot theory and its
applications.
300
0196-8858r97 $25.00
Copyright Q 1997 by Academic Press
All rights of reproduction in any form reserved.
 RATIONAL TANGLES 301

The paper is organized as follows. Section 1 defines the notions of

tangle, rational tangle, and the continued fraction and fraction associated
with a rational tangle. Then, by using a few simple lemmas about the
topology of rational tangles coupled with an elementary algebraic identity
of Lagrange for continued fractions, we show that tangles ha¨ ing the same
fraction are ambient isotopic. Our proof actually provides an algorithm for
transforming a rational tangle into an equivalent one. This is one half of
the ‘‘if and only if’’ of the Conway Theorem. What is unique in our
approach to this half is the fact that a topological interpretation for
Lagrange’s identity provides the simple key to the proof. This completes
the ‘‘hard’’ part of the Theorem and Section 3 turns to the ‘‘easier’’ part of
showing that the fraction of a rational tangle is a topological invariant of
that tangle. That is, we must show that if two tangles are ambient isotopic,
then they have the same fraction. This task is accomplished by showing
that a topologically invariant ‘‘fraction’’ can be defined for any tangle
whatsoever, and that this fraction agrees with the rational tangle fraction
that we have defined in Section 2. We define the general tangle fraction by
using the bracket model of the Jones polynomial w8, 9x at a special value.
The advantage of this approach is that the bracket properties can be
verified by entirely elementary means, and the bracket formalism is just
suited to handling the tangles.
We also remark that the definition of tangle fraction given here coin-
cides with the definition of conductivity for tangles explained in our earlier
paper w7x. In fact, we can use the bracket definition of conductivity to
deduce the formula for the conductivity of a dual graph in the plane
ŽRemark after Theorem 3.3.. The discussion in Section 3 completes our
proof of the Conway Theorem. Section 4 is a quick discussion of the
applications of rational tangles to the analysis of DNA recombination.

RATIONAL TANGLES AND THEIR ALGEBRA}

CONWAY’S THEOREM

The key concept for this section is the notion of a tangle. A tangle is
analogous to a link except that it has free ends. These ends are depicted as
strands that enter a box Žthe tangle box. within which there are no free
ends. Inside the box one may find closed loops that are knotted and linked
with the tangle strands. The strands of the tangle may themselves be
knotted and linked. Note that if you begin at one of the free ends
emanating from the tangle box and walk along it, you will enter the box,
and eventually leave the box to meet another end of strand. Thus a tangle
box must have an even number of ends. We will be concerned in this
section with tangles with four ends. Such tangles have two strands, each
302 GOLDMAN AND KAUFFMAN

strand participating in two ends. In considering the topology of tangles, we

allow the strands to move inside the box by ambient isotopy, but the ends
of the strands must remain fixed, and the strand movements are confined
to the box.
Rational tangles were introduced by John Conway as basic building
blocks for the construction of knots. They are the simplest tangles in the
sense that they can be ‘‘unwound.’’ More precisely, we visualize the four
endpoints of the strands of a tangle as restricted to the surface of a sphere
and the rest of the tangle inside the sphere. If we restrict the endpoints of
the tangle to move on the surface and the rest of the tangle to move inside
the sphere, then a rational tangle is one that can be deformed into two
straight lines. In Fig. 2.1A, the tangle can be unwound by grabbing the
ends b and c and untwisting while holding a and d fixed. In Fig. 2.1B, we
unwind by first using the ends c and d to unwind the bottom and then
using b and d to unwind the remaining twist.
Conversely, we can construct a rational tangle by starting with two
horizontal or two vertical strands, picking two endpoints and twisting them,
then picking another pair and twisting them, and so on, for a finite number
of twists. We see an example in Fig. 2.1B by reading the figures from the
bottom up. A twist of two diagonally opposite strands can be obtained by
two successive twists of adjacent strands ŽFig. 2.2.. Thus we only need
twists of adjacent structure to construct rational tangles.
We need a few definitions before defining a rational tangle formally.
A horizontal Žresp. ¨ ertical . integer tangle t a Žresp., tXa . is a twist of two
horizontal Žresp. vertical. strands < a < times in the positive or negative
direction according to the sign of a. The directions are shown in Fig. 2.3.
The horizontal sum ‘‘q’’ and ¨ ertical sum ‘‘q9 ’’ of two tangles A and B
are defined by the diagrams in Fig. 2.4. Thus twisting two adjacent strands
of a tangle A is equivalent to adding an integer t a on the right or the left
Žwith q. or adding tXa on the top or bottom Žwith q9. ŽFig. 2.5..
It is easy to see that t a q t b ; t aqb and tXa q9tXb ; tXaqb ŽFig. 2.6..
Essentially, a negative twist cancels a positive twist topologically so that
cancellation of positive and negative integers is paralleled in tangle topol-
ogy. This particular cancellation is an instance of the second Reidemeister
move in knot theory. See Section 3 of this paper for a discussion and
illustration of the Reidemeister moves.
At this point it is worth drawing attention to the notion of ambient
isotopy of tangles. Two tangles A and B are said to be ambient isotopic if
it is possible to deform A into B without mo¨ ing the endpoints and without
mo¨ ing the strands outside the tangle box Žwe will usually just say isotopic..
When we say deform, we mean topological deformation, a movement of
the strands that is continuous and does not allow any strand to penetrate
 RATIONAL TANGLES 303

FIGURE 2.1

either itself or another strand. Of course, we usually visualize this by

taking a representation of A and a representation of B, and trying to
change A until it looks like B. In the end, if A and B are isotopic, the
changed version of A should be identical to B. In Section 3 we will explain
how isotopy can be expressed by the combinatorial Reidemeister moves on
diagrams. It will then be possible to give ways to calculate invariants of
isotopy. In this section, we use some elementary isotopies to simplify the
presentation of rational tangles.
304 GOLDMAN AND KAUFFMAN

FIG. 2.2. Twining diagonally opposite strands via twists of adjacent strands.

Now we define a rational tangle Žwith n integer tangles. inductively.

DEFINITION. For any sequence of integers a1 , . . . , a n , choose a se-
quence of integer tangles Ta1, . . . , Ta n, where Ta i s t a i or tXa i .
Ž1. Let B1 s Ta .
1

Ž2. For k - n.
Ži. if Ta s t a , let Bkq1 s t a q Bk or Bk q t a ,
kq 1 kq 1 kq 1 kq 1

Žii. if Ta s tXa , let Bkq1 s tXa q9Bk or Bk q9tXa .

kq 1 kq 1 kq 1 kq 1

Any tangle constructed in this way is a rational tangle with n integer

tangles.
Note. The number of integer tangles in a rational tangle is not unique.
For example, t5 ; t 3 q t 2 , and, by our definition, t5 has one integer tangle
and t 3 q t 2 has two.
Note. This definition of ‘‘rational tangle with n integer tangles’’ is
simply a formalization of our description of a rational tangle as obtained
from two horizontal or two vertical strands by successive twisting of pairs
of adjacent endpoints. These twistings can be construed as the Žhorizontal
 RATIONAL TANGLES 305

FIGURE 2.3

or vertical. addition of horizontal or vertical tangles on the right, left, top,

or bottom. Thus any rational tangle is a rational tangle with n integer tangles
for some n. The next theorem gives a fundamental simplification in the
description of rational tangles.

FLIP THEOREM 1. A 1808 rotation Ž flip . of a rational tangle b in the

horizontal or ¨ ertical axis is ambient isotopic to b.
306 GOLDMAN AND KAUFFMAN

FIGURE 2.4

Note. This theorem is, in general, not true for nonrational tangles. For
example, consider the tangle illustrated in Figure 2.6.1. It is not rational.
One strand has a knot in it and this strand is linked with the other strand.
This tangle is not equivalent to the tangle obtained by performing a
vertical flip because there is no way to deform the knot on one strand to
the other one, since the endpoints are fixed.

Proof of Flip Theorem 1. We proceed by induction on the number n of

integer tangles in b.
If n s 1, then b s t a or tXa for some a. In either case a horizontal or
vertical flip leaves the tangle invariant.
Assume the theorem is true for any rational tangle with F n integer
tangles, and let b be a tangle with n q 1 integer tangles. By the definition
of tangles b is equal to one of the tangles t a q B, B q t a , tXa q9B, or
B q9tXa for some a and some rational tangle B with n integer tangles.
Suppose that b s t a q B. Then a horizontal flip leaves the t a part
invariant and, by the induction hypothesis, takes B into a tangle B n
isotopic to B. ŽNote that, as discussed above, tangle isotopy leaves the
endpoints of the tangle fixed, so that the isotopy of B n to B induces an
isotopy of t a q B n to t a q B.. Hence t a q B is isotopic to its horizontal
flip t a q B n.
 RATIONAL TANGLES 307

FIGURE 2.5

Vertical flips take a little more work. Consider specifically b s t 3 q B.

Its vertical flip is B9 q t 3 , where B9 is the vertical flip of B. By the
induction hypothesis B ; B9, so B9 q t 3 ; B q t 3 . Hence we must prove
that t 3 q B ; B q t 3 . To see this, we twist B through three 1808 twists
Žholding the endpoints of t 3 q B fixed., as shown in Fig. 2.7. Such a flip of
a subtangle is called a flype. Since we performed an odd number of flypes,
the resulting subtangle becomes B n Žthe horizontal flip of B .. Again, by
the induction hypothesis, B n; B and we are done.
The other cases are handled in exactly the same way, using horizontal or
vertical flypes as needed. The key idea is that flyping allows us to mo¨ e an
integer tangle from the right to the left, from the left to the right, from the
bottom to the top, and from the top to the bottom.
We state an essential part of the proof separately.

COROLLARY. If B is a rational tangle and a is an integer, then

ta q B ; B q ta and tXa q9B ; B q9tXa .

308 GOLDMAN AND KAUFFMAN

FIGURE 2.6
RATIONAL TANGLES 309

FIGURE 2.6.1

FIGURE 2.7
310 GOLDMAN AND KAUFFMAN

Now we define a subclass of rational tangles, the basic tangles. A basic

horizontal Žresp., ¨ ertical . tangle is a rational tangle built as follows:
Ž1. Start with a horizontal tangle t a Žresp., a vertical tangle tXa ..
Ž2. Add a vertical tangle tXb on the bottom Žresp., a horizontal tangle
t a on the left ., then a horizontal tangle t c on the right Žresp., a vertical
tangle tXc on the bottom., then a tXd on the bottom Žresp., t d on the left.,
and so on, stopping after a finite number of such steps.
If we start with t a , the resulting tangle is a basic horizontal tangle, and if
we start with tXa , it is a basic vertical angle.

For example, the construction t 3 ª t 3 q9tX2 ª Ž t 3 q9tX2 . q ty2 ends in a

basic horizontal tangle, whereas tX3 ª t 2 q tX3 ª Ž t 2 q tX3 . q9ty2 ends in a
basic vertical tangle ŽFig. 2.8..
When constructing a rational tangle, we can use the last corollary to
convert horizontal twists on the left to those on the right and vertical twists
on the top to those on the bottom.
Moreover, the corollary plus the fact that t a q t b ; t aqb and tXa q9tXb ;
X
t aqb allows us to avoid adding two horizontal or two vertical twists in a
row. Hence we have
THEOREM 2. E¨ ery rational tangle is isotopic to a basic tangle.
Note that an algorithm is implicit in the proof of Theorem 2.
In order to classify the basic tangles, we introduce three more opera-
tions.

FIGURE 2.8
 RATIONAL TANGLES 311

DEFINITION. If b is a rational tangle, then

Ž1. yb s b* is the mirror image of b Žreverse all crossings.

Ž2. The c-in¨ erse Žclockwise inverse. 1r c b of b is the tangle ob-

tained by rotating b 908 clockwise and taking the mirror image.

Ž3. The cc-in¨ erse Žcounterclockwise inverse. 1rcc b of b is the

tangle obtained by rotating b 908 degrees counterclockwise and taking the
mirror image.
312 GOLDMAN AND KAUFFMAN

FIGURE 2.9

The reason for taking mirror images in the definition of inverses will
become clear when we discuss numerical invariants.
THEOREM 3. The c-in¨ erse and cc-in¨ erse of a rational tangle are ambient
isotopic.
Proof. We apply the flip theorem twice Ža vertical followed by a
horizontal flip as shown in Fig. 2.9.
Thus if a subtangle of a tangle is of the form 1r c b, b rational, we can
always replace it by 1rcc b and, conversely, we can replace 1rcc b by 1rc b.
We will often write 1rb if the type of the inverse does not matter.
We need a few more properties of our operations in order to describe
basic tangles algebraically.
PROPOSITION 4. Ž1. If b is a basic horizontal tangle, then 1r c b is a basic
¨ ertical tangle.
Ž2. If b is a basic ¨ ertical tangle, then 1rcc b is a basic horizontal tangle.
Ž3. tXa s 1rc t a s 1rcc t a and t a s 1rc tXa s 1rcc tXa .
Ž4. t d q tXe s t d q 1rt e .
1
Ž5. t d q9tXe s cc.
1
te q c
td
Ž6. yŽ b q c . s Žyb . q Žyc ., yŽ1rb . s 1rŽyb ., yt a s tya , and
ytXa s tya
X
.
Proof. The results follow easily from the definitions Žwe leave it to the
reader to draw these pictures..

Now we can build our basic tangles using the tangle operations q, y,
r c , rcc . For example, consider the basic tangles of Fig. 2.8. The construc-
tion of the first tangle is given by

t 3 ª t 3 q 9t 2 ª Ž t 3 q 9tX2 . q ty2
 RATIONAL TANGLES 313

or
1 1
t3 ª cc ª cc q ty2
1 1
t2 q c t2 q c
t3 t3
1
; ty2 q cc
1
t2 q c
t3

and the second is given by

tX3 ª t 2 q tX3 ª Ž t 2 q tX3 . q9ty2
X

or
1 1 1
ª t2 q ª cc
t3 t3 1
ty2 q c
1
t2 q
t3

In both cases we have a ‘‘continued fraction’’ in elementary tangles.

Moreover, the second tangle is isotopic to the reciprocal of the first Žrecall
Theorem 3..
Similarly, any basic horizontal tangle b can be written in the form
1
b s tan q Ž 1.
1
t a ny 1 q .
..
1
q
t a1

and any basic vertical tangle b in the form

1
bs . Ž 2.
1
tan q
1
t a ny 1 q .
..
1
q
t a1

In fact, we could take Ž1. and Ž2. as the definition of a basic tangle.
314 GOLDMAN AND KAUFFMAN

These forms for basic tangles suggest associating to the basic tangles Ž1.
and Ž2. the arithmetic continued fractions
1 1
an q and . Ž 3.
1 1
a ny 1 q . an q
.. 1
a ny 1 q .
1 ..
q
a1 1
q
a1

Each of these continued fractions equals a rational fraction which we call

the fraction of the tangle and denote it by F Ž b ..
In particular, F Ž t a . s a and F Ž tXa . s F Ž1rt a . s 1ra.
PROPOSITION 5. If b is a basic tangle, then
Ž1. F Ž1rb . s 1rF Ž b ., independent of which in¨ erse we use.
Ž2. If t a q b is also a basic tangle, then F Ž t a q b . s F Ž t a . q F Ž b . s
a q F Ž b ..
Ž3. F Žyt a . s F Ž tya . s ya s yF Ž t a ..

F Ž ytXa . s F Ž tya
X
. s F Ž 1rtya . s y1ra s yF Ž 1rt a . s yF Ž tXa . .

Ž4. F Žyb . s yF Ž b ..
Proof. Only Ž4. is not immediate from the definitions and Proposi-
tion 4. A formal proof of Ž4. requires induction, but we shall just illustrate
the reasoning with an example. Let

1
b s t3 q .
1
t4 q
ty5

Then, by Proposition 4,
1
yb s yt 3 y
1
t4 q
ty5
1
s yt 3 q
1
yt4 y
ty5
 RATIONAL TANGLES 315

1
s yt 3 q
1
yt4 q
yty5
1
s yt 3 q .
1
yt4 q
t5
Therefore

1 1
F Ž yb . s y3 q sy 3q s yF Ž b . .
1 1
y4 q 4q
5 y5
We now have the concepts and tools to state and prove John Conway’s
fundamental theorem.
CONWAY’S THEOREM. Let T1 and T2 be basic tangles. If F ŽT1 . s F ŽT2 .,
then T1 is ambient isotopic to T2 .
Remarks. The converse is in fact true, namely, if T1 is ambient isotopic
to T2 , then F ŽT1 . s F ŽT2 .. Later we will give a different interpretation of
the fraction of a tangle, which allows us to prove this. Thus the fraction is
a complete invariant for the equivalent of basic tangles. Moreover, since
any rational tangle is ambient isotopic to a basic tangle Žby an algorithm.,
we have an algorithmic procedure for deciding the equivalence of any two
rational tangles.
Proof. Given a basic tangle T in the continued fraction form of Eq. Ž1.
with fraction prq, we will show that it is ambient isotopic to a basic tangle
T with fraction prq, whose continued fraction Ž3. satisfies a i ) 0 for
i s 1, . . . , n y 1. Such a continued fraction is called regular. Since the
regular continued fraction of prq is unique Žup to the last term which can
be written as a n y 1 q 1r1}see w5x., this means that any tangle with
fraction prq is ambient isotopic to T 9 and we will be finished.
We begin arithmetically with the continued fraction Ž3. of the fraction
F ŽT . and show how to convert it into the regular continued fraction with
the same sum. The key step is the following formula of Lagrange ŽSee
Lagrange’s Appendix to Euler’s Algebra w5x.,
1 1
ay s Ž a y 1. q , Ž 4.
b 1
1q
Ž b y 1.
which is trivial to verify.
316 GOLDMAN AND KAUFFMAN

Applying Lagrange’s formula to convert a continued fraction to a

regular one is best illustrated with an example. Consider the continued
fraction

43 1 1
s1q s1y Ž 5.
62 1 1
y3 q 3q
1 1
y4 q 4y
5 5

and move all negative signs to the numerator as shown. We find the last
appearance of a negative sign reading down the fraction Žin this case
y1r5. and apply Eq. Ž4. to 4 y 1r5 Ž a s 4, b s 5 y 1r5.. Then

1 1
4y s3q ,
5 1
1q
4

which, substituted into the right-hand side of Ž5. yields

43 1
s1y , Ž 6.
62 1
3q
1
3q
1
1q
4

a continued fraction for 43r62 with one less negative sign. Now find the
last negative sign in Ž6., which is the only one, and apply Eq. Ž4. to Ž6. with
a s 1 and
1
bs3q ,
1
3q
1
1q
4

which gives

43 1 1 1
say s0q s ,
62 b 1 1
1q 1q
1 1
2q 2q
1 1
1q 1q
4 4
 RATIONAL TANGLES 317

a continued fraction with all terms positive. This procedure, namely using
Lagrange’s formula to reduce the number of negative signs one step at a
time, works in general yielding a continued fraction in which a negative
sign can only appear in front of the first term.
Now we mirror this procedure topologically. Let the numbers a and b
represent rational tangles with fractions a and b respectively. Then
Fig. 2.10 shows the construction of a rational tangle with fraction 1rŽ1 q
1rŽ b y 1.., which corresponds to part of the right-hand side of Ž4..

FIGURE 2.10
318 GOLDMAN AND KAUFFMAN

In Fig. 2.11, we use the rational tangle from Fig. 2.10 to form a rational
tangle corresponding to Ž a y 1. q 1rŽ1 q 1rŽ b y 1.. and then topologi-
cally deform this tangle until it is the rational tangle for the fraction
a q 1rŽyb .. Thus we have the topological version of Lagrange’s formula.
Now we are essentially done. We start with a basic tangle T in the
continued fraction form of Eq. Ž1. with fraction prq given by the contin-
ued fraction Ž3.. We have shown how to convert the continued fraction
into a regular one, which is essentially unique. Then we showed how to
convert the original tangle to an ambient isotopic tangle corresponding to
the regular continued fraction for prq. Hence given any two tangles T1
and T2 with the same fraction prq, they are ambient isotopic to the same
tangle T 9 which corresponds to the unique regular continued fraction for
prq. Therefore T1 and T2 are ambient isotopic. Moreover, our proof
actually gives an algorithm for converting T1 to T2 .
Note that the basic tangles associated to regular continued fractions are
alternating, i.e., as we walk along a strand, the crossings alternate between
under and over crossings. So we have proved that any basic tangle is
ambient isotopic to an alternating tangle.

FIGURE 2.11
 RATIONAL TANGLES 319

3. THE CONDUCTANCE INVARIANT AND

THE BRACKET

We have seen how to define the fraction of a rational tangle and that
tangles with the same fraction are ambient isotopic. In this section we shall
prove the converse}that if two tangles are ambient isotopic, then they
have the same fraction. This will complete our goal of showing that the
fraction of a rational tangle completely classifies its topological type.
In order to show that ambient isotopic tangles have the same fraction,
we shall build an in¨ ariant, C ŽT ., for arbitrary tangles T and then show
1. If T is ambient isotopic to S, then C ŽT . s C Ž S ..
2. If T is rational, then C ŽT . s F ŽT ., the fraction of T.
This is a case where it is actually easier to solve a more general problem
Žof finding an invariant for arbitrary tangles rather than just for rational
tangles..
We shall call the invariant C ŽT . the conductance of T, because it is the
Žgeneralized. conductance of a Žgeneralized. electrical network associated
with T. The full theory of C ŽT . is explained in our paper w7x, but here we
shall give a different and elementary development of C ŽT . that is based on
the bracket polynomial. In order to do this we will first give a quick
introduction to the bracket. Readers interested in more information on the
bracket polynomial should consult w8, 9, or 11x. The relationship between
the bracket ŽJones polynomial. and the conductance invariant is explained
in w7x.

The Bracket
The bracket invariant ² K : is defined on all knots and links K Žnot yet
on tangles. by the formulas:
Ž1.

Ž2. ² OK : s d ² K :,
² O : s 1,
320 GOLDMAN AND KAUFFMAN

where d s yA2 y Ay2 and B s Ay1 . The assignment of A and B s Ay1

in Ž1. follows the labeling convention}the regions swept out as the o¨ er
crossing strand is rotated counterclockwise to the under crossing strand are
labeled with A; the other regions are labeled with Ay1 . The symbol A is thus
associated with splicing the crossing from the A to the A regions and B
with splicing the crossing from the Ay1 to the Ay1 regions. See w8x for the
motivation behind this definition. In the formula Ž1. the small diagrams
stand for otherwise identical parts of larger diagrams. In formula Ž2. it is
stated that the appearance of an extra loop Ž O . multiplies the bracket by
d s yA2 y Ay2 , and that the solitary loop receives the value 1. ŽIn other
words, ² : s 1rd . The unmarked plane has bracket value 1rd ..
These rules give a well-defined polynomial ² K : in A and Ay1 associ-
ated to a given link diagram K Žsee w8, 9, 11x and Exercise 0 below..

Exercises in Bracketology
The following exercises provide a self-contained introduction to various
basic notions and to the bracket polynomial. Some of these results will be
used later.

0. INFORMATION ON THE REIDEMEISTER MOVES. In w10x Reidemeister

proved that the set of two-dimensional moves on diagrams shown in Fig.
3.1 are sufficient to capture the concept of ambient isotopy of knots and
links in three-dimensional space. That is, two knots and links are ambient
isotopic if and only if diagrammatic ‘‘snapshots’’ of each Žprojections on a
plane where at most two strands meet at a crossing. are equivalent by a
sequence of Reidemeister moves combined with homeomorphisms of the
diagrams in the plane that do not change the crossing structure.
For an exercise, unknot the knot shown in Fig. 3.2 using only Reideme-
siter moves. For a second exercise, turn the figure-eight knot in Fig. 3.2
into its mirror image via the Reidemeister moves. For the latter, it may
help to make a model of the figure eight from rope and see by direct
topological manipulation that the figure-eight knot is indeed equivalent to
its mirror image.

1. Let ² K : denote the bracket polynomial in independent commuting

variables A, B, and d . ŽWe do not yet assume that B s Ay1 or that
d s yA2 y Ay2 .. Verify that the recursion formulas Ž1. and Ž2. in the
definition of the bracket polynomial yield a well-defined polynomial func-
tion of link diagrams in the three variables A, B, and d .

2. Continue with the assumptions of problem 1 and show that the

three-variable bracket polynomial has the following behavior under Reide-
 RATIONAL TANGLES 321

FIG. 3.1. Reidemeister moves.

FIGURE 3.2
322 GOLDMAN AND KAUFFMAN

meister move II:

It follows from this formula that if we take B s Ay1 and d s yA2 y Ay2 ,
then

With this specialization of the variables, the bracket is invariant under the
second Reidemeister move.
3. Show that with B s Ay1 and d s yA2 y Ay2 , the invariance of the
bracket under the second Reidemeister move implies its invariance under
the third Reidemeister move.
4. Show that ² K : changes under the first Reidemeister move via the
formulas
 RATIONAL TANGLES 323

Note that in working with the bracket calculations we are not allowed to
perform type I moves on the diagrams without applying these compensat-
ing formulas.
5. Show that if K * is the mirror image of K, obtained by switching all
the crossings in the diagram for K, then ² K *:Ž A. s ² K :Ž Ay1 . where we
now use the bracket with B s Ay1 and d s yA2 y A2 . Compute the
bracket polynomial in A and Ay1 for the trefoil diagram in Fig. 3.3.
6. Let K be an oriented link diagram and define the writhe of K, w Ž K .,
by the formula w Ž K . s the sum of q1 or y1 for each crossing in K
where the q1 or y1 is the sign of the crossing.
The sign of a crossing is obtained as shown in Fig. 3.4.
Now define the normalized bracket, f K Ž A., by the formula f K Ž A. s
ŽyA3 .yw Ž K . ² K : where we take K to be oriented but forget the orienta-
tion when we compute the bracket. Show that f K Ž A. is an invariant of all
three Reidemeister moves. Show that f K * Ž A. s f K Ž Ay1 .. Use these facts
and your calculation of the trefoil diagram Žfrom Problem 4. to prove that
the trefoil knot is chiral Ži.e., inequivalent to its mirror image..

This ends the exercises in bracketology.

We now introduce the bracket for tangles.
DEFINITION. Let T be a tangle. Then
²T : s a Ž T . ²5: q b Ž T . ² s : ,
where the coefficients a ŽT . and b ŽT . are obtained by starting with T and
using formulas Ž1. and Ž2. repeatedly until only the infinity and zero
tangles are left. We let w0x denote the 0-tangle and w`x denote the infinity
tangle so that ²T : s a ŽT .²w`x: q b ŽT .²w0x:. Neither ²w0x: nor ²w`x: are
numbers. They are place holders for the results of the computation of the
bracket restricted to the tangle.

FIGURE 3.3
324 GOLDMAN AND KAUFFMAN

FIGURE 3.4

Figure 3.5 shows an example of a bracket calculation.

THEOREM 3.1. For any choice of A, R T Ž A. s a ŽT .rb ŽT . is an ambient
isotopy in¨ ariant of tangles.
Proof. The main point is to examine how a ŽT . and b ŽT . behave under
the Reidemeister moves. As outlined in the exercises above, one can
expand the tangles in the same way that one expanded a knot or a link to
calculate the bracket. In so doing, the verification of invariance under the
second and third Reidemeister moves goes through in the same way as for
links}with the caveat that one keep the ends of the tangle fixed and that

FIGURE 3.5
 RATIONAL TANGLES 325

only isotopies within the tangle box are allowed. By the properties of the
bracket expansion, a ŽT . and b ŽT . are each invariant under Reidemeister
II and Reidemeister III moves of the tangle. We know that the bracket
behaves under a type-I move by multiplying by either yA3 or by yAy3 . It
is easy to see that both a ŽT . and b ŽT . are multiplied by the same factor
under a type-one move applied to the tangle. Hence the ratio R T Ž A. s
a ŽT .rb ŽT . is unchanged under the first Reidemeister move. Therefore
the ratio R T Ž A. is an invariant of the tangle T. Note that RŽT . may take
the value ‘‘infinity’’ under the circumstance that b ŽT . is zero when a ŽT .
is nonzero. This completes the proof of the theorem.

Specializing the A
Now let A s 'i Ž i 2 s y1.; hence B s 1r 'i and d s yA2 y Ay2 s
yi y 1ri s yi q i s 0. Thus the loop value is 0 for this bracket. From
now on ²T : denotes this specialization of the bracket. Let
C Ž T . s yiR T Ž 'i . .
We are now going to see that C ŽT . has just the properties that we need to
show that C ŽT . s F ŽT . for rational tangles.
It is easy to see that C Ž1rT . s Ž1rC ŽT ..*, where * denotes complex
conjugation, for
²T : s a Ž T . ² w ` x : q b Ž T . ² w 0 x :
« ²1rT : s a Ž T . *² w 0 x : q b Ž T . *² w ` x : ,
since
1r w 0 x s w ` x as tangles and Ž 1r'i . * s 'i .
Thus C Ž1rT . s b ŽT .*ri a ŽT .* s i*b ŽT .*ra ŽT .* s Ž1rC ŽT ..*.
Note that when C ŽT . is real, then C Ž1rT . s 1rC ŽT . since a real
number is conjugate to itself. We shall see shortly that the values of C ŽT .
on rational tangles are rational numbers and hence real.
Thus for rational tangles, C Ž1rT . s 1rC ŽT ..
EXAMPLE.

s 'i ²5: q Ž 1r'i . ² s : s 'i ² w ` x : q Ž 1r'i . ² w 0 x : .

326 GOLDMAN AND KAUFFMAN

Hence C Žw1x. s Ž1ri .Ž'i rŽ1r 'i .. s Ž1ri .Ž ir1. s 1. In exactly the same
way, we find that C Žwy1x. s y1. Note that C Žw`x. s 0 and cŽw0x. s `,
with the convention that formally 1r0 s `.
We now see that C ŽT q S . s C ŽT . q C Ž S ..
PROPOSITION 3.2. If ²T : s a²w`x: q b ²w0x: and ² S : s c²w`x: q d²w0x:
then ²T q S : s Ž ad q bc .²w`x: q bd²w0x:. Hence C ŽT q S . s C ŽT . q
C Ž S . follows from the fact that Ž ad q bc .rbd s arb q crd.
Proof. In this proof we shall proceed by a picture-writing technique
ŽFig. 3.6.: the boxes stand for the tangles in question. We first expand on
the tangle T, replacing it by the sum of two tangles with coefficients that
corresponds to its bracket expansion. We then expand each of these
pictures on the tangle S to get the full sum that is evaluated. Thus
²T q S : s Ž ad q bc .²w`x: q bd²w0x:, completing the proof of the theo-
rem.

This completes the proof of all the properties of C ŽT . that we need.

Since the generating tangles w1x and wy1x take rational values under C ŽT .
it follows that tangles generated from them will also take Žreal. rational
values so that C Ž1rT . s 1rC ŽT . for any rational tangle. Then repeated
application of this reciprocal formula, coupled with the addition theorem
C ŽT q S . s C ŽT . q C Ž S ., implies that C ŽT . s F ŽT . for rational tangles.

FIGURE 3.6
 RATIONAL TANGLES 327

Our mission is complete. We have proved that the rational tangles are
classified by their fractions.

Remark on Numerators and Denominators. To any tangle T there are

associated two links, the numerator of T, N ŽT ., and the denominator of
T, DŽT .. The numerator is obtained by tying the input strands to each
other and the output strands to each other ŽFig. 3.7.. The denominator is
obtained by tying the input strands to the output strands as indicated in
Fig. 3.7. It is easy to see that the conductance, C ŽT ., of the tangle is i
times the ratio of the bracket evaluations of the numerator and the
denominator of the tangle. That is,

THEOREM 3.3. C ŽT . s yi ² N ŽT .:r² DŽT .:.

Proof. When ²T : s a²w`x: q b ²w0x:, it follows that ² N ŽT .: s a, while

² DŽT .: s b. Thus C ŽT . s yiarb s yi ² N ŽT .:r² DŽT .:.

In the next section we shall have occasion to use numerators and

denominators of tangles associated with DNA recombination.

Remark on Graphs and Duals. Before leaving this topic there is an

observation about the usefulness of the bracket that is worth making. In
order to make this remark we will assume the background of our paper w7x
without further explanation. In that paper we associate to the tangle T a
signed plane graph called GŽT, ¨ , ¨ 9. with special vertices ¨ and ¨ 9
corresponding to the inputs and outputs of the tangle. We regard this
signed plane graph as a generalized electrical network. We prove that
C ŽT . is equal to the Žgeneralized. conductance of the network between the
vertices ¨ and ¨ 9 where each edge has conductance "1 according to its
sign and negative conductances are handled algebraically in the same

FIG. 3.7. Numerator and denominator.

328 GOLDMAN AND KAUFFMAN

manner as classical positive conductances in a linear circuit. Let T * denote

any choice of inverse tangle for T. It follows that the graph GŽ1rT, w, w9.
is the planar dual to the graph GŽT, ¨ , ¨ 9. and, by using the bracket, we
conclude that the conductance of the graph is the inverse of the conduc-
tance of its planar dual. This result comes directly from our easy proof that
C Ž1rT . s 1rC ŽT .. It is not at all obvious how to produce such a direct
argument using pure graph combinatorics. This shows how the bracket
model for conductance is not just a trick, but in fact it is a way to see more
deeply the properties of both the topology and the electricity associated
with plane graphs.

IV. MOLECULAR BIOLOGY

In this section we sketch an application of fractions of rational tangles to

molecular biology. This method of using tangles has been pioneered by
DeWitt Sumners w14x and used in the work of Cozzarelli and Spengler w4x.
See Kauffman w11, Part 2x and Adams w1x for introductions to the subject.
Recombination of DNA is the process of cutting two neighboring
strands with an enzyme and then reconnecting them in a different way.
The idea of applying tangle theory is to use the addition of tangles to write
the equations for possible recombinations of DNA molecules. Then one
uses topological information Žsuch as the fraction of tangle. to obtain
limitations on the possibilities for the products of the recombination.
Recombination occurs in successive rounds for which the nature of the
products can be known through a combination of electrophoresis and
electron microscopy. In particular, electron microscopy provides the biolo-
gist with an enhanced image of the DNA molecule from which it is
possible to see direct evidence of knotting and supercoiling. In the case of
TN3 resolvase, a species of closed circular DNA is seen to produce very
specific knots and links in successive rounds of recombination. By knowing
these actual products of the rounds of recombination it is possible to use
topology to deduce the mechanism for the recombination.
In order to apply the fraction of a tangle to molecular biology, we shall
make the blanket assumption that all products of recombination, starting
from a gi¨ en unknotted and unlinked form of double-stranded DNA, are
closures (numerators) of rational tangles. This is a reasonable assumption. It
assumes that the knots or links that are built in the recombination process
are obtained by a combination of simple twisting Žof the sort that builds
new rational tangle from old. and the addition of single crossings at a
smoothing site. The latter operation is what is usually called site specific
recombination by biologists Žsee Fig. 4.1.. A crossing is created in place of
the smoothing that is the local configuration of the ‘‘lined-up’’ sites. There
 RATIONAL TANGLES 329

FIG. 4.1. Site specific recombination.

are two possibilities for such a crossing. We have called these possibilities
R and L. In Fig. 4.2 we have illustrated the concept of a site specific
recombination by drawing the DNA schematically in a single closed loop
with local arrows at these sites. In order for the recombination to occur,
the DNA must twist about to bring these two sites into proximity with the
orientations lined up.

FIG. 4.2. First round of recombination.

330 GOLDMAN AND KAUFFMAN

We regard the crossing R as a small tangle and differentiate it from its

reverse version L. Now view Fig. 4.2. Here we see the result of a single
round of recombination. First it is assumed that there is a total twist of the
DNA and that the two sites are brought into proximity. In this form, the
DNA can be described as the numerator, NumŽw1rn x q U ., of the tangle
sum w1rn x q U where U denotes the tangle that we have previously called
w0x. The recombination then produces NumŽw1rn x q S ., where S equals R
or L. In Fig. 4.2, we have taken n s q3 and S s L.
Successive rounds of recombination produce

Num Ž w 1rn x q S q S . ,
Num Ž w 1rn x q S q S q S . ,
??? .

For example, let n s 3 and S s L. Then the successive rounds of

recombination are shown in Figs. 4.3 and 4.4. The first two rounds give a
simple link and the figure-eight knot. The third round gives a link of two
components with linking number zero. In fact, TN3 resolvase produces just
these knots and links in its successive rounds. Here we have indicated a
possible mechanism for TN3 resolvase. Is it the only possibility?
In order to answer this question in the context of our model, we
consider the most general case of a sequence of rational tangles in the
form T, T q S, T q S q S, . . . such that FRACŽT q S . s 1r3 y 1 s
y2r3 and FRACŽT q S q S . s 1r3 y 1 y 1 s y5r3. Let x s
FRACŽT . and y s FRACŽ S .. Then Žusing the fact that the fraction of a

FIG. 4.3. Second round of recombination.

 RATIONAL TANGLES 331

FIG. 4.4. Third round of recombination.

sum of tangles is the sum of the fractions of the summands. we have the
equations

x q y s y2r3
x q 2 y s y5r3.

The only solution to these equations is x s 1r3 and y s y1. By the

Conway Fraction Theorem, the tangles T and S are identified as T s w1r3x
and S s wy1x. This shows how the topology can be used to pinpoint a
biological mechanism.

ACKNOWLEDGMENTS

It gives Lou Kauffman great pleasure to thank John Conway for numerous conversations
about tangles and square dancing and David Krebes for insightful remarks on tangles and the
bracket. We both thank DeWitt Sumners for his enthusiasm for the topology of DNA and
Steve Bleiler for helpful conversations.

REFERENCES

1. C. Adams, ‘‘The Knot Book,’’ Freeman, San Francisco, 1994.

2. G. Burde and H. Zieschang, ‘‘Knots,’’ de Gruyter, Berlin, 1986.
3. J. H. Conway, An enumeration of knots and links and some of their algebraic properties,
in ‘‘Computational Problems in Abstract Algebra,’’ ŽD. Welsh, Ed.., pp. 329]358, Perga-
mon Press, New York, 1970.
332 GOLDMAN AND KAUFFMAN

4. N. R. Cozzarelli, S. J. Spengler, and A. Stasiak, The stereostructure of knots and

catenanes produced by phase l integrative recombination: implications for mechanism
and DNA structure, Cell 42 Ž1985., 325]334.
5. H. Davenport, ‘‘The Higher Arithmetic: An Introduction to the Theory of Numbers,’’
Hutchinson University Library, London, 1952; reprint, Dover, New York.
6. L. Euler, ‘‘Elements of Algebra,’’ 5th Ed., ŽJ. Hewlett. Longman, Orme, and Co., London,
1840; reprint Springer-Verlag, New YorkrBerlin, 1984.
7. J. R. Goldman and L. H. Kauffman, Knots tangles and electrical networks, Ad¨ . Appl.
Math. 14 Ž1993., 267]306.
8. L. H. Kauffman, State models and the Jones polynomial, Topology 26 Ž1987., 395]407.
9. L. H. Kauffman, New invariants in the theory of knots, Amer. Math. Monthly 95, No. 3
ŽMar. 1988., 195]242.
10. L. H. Kauffman, ‘‘On Knots,’’ Annals of Mathematics Studies, Vol. 115, Princeton Univ.
Press, Princeton, NJ, 1987.
11. L. H. Kauffman, ‘‘Knots and Physics,’’ World Science, Singapore, 1991 and 1994.
12. L. H. Kauffman and S. L. Lins, ‘‘Temperley Lieb Recoupling Theory and Invariants of
3-Manifolds,’’ Annals of Mathematics Studies, Vol. 134, Princeton Univ. Press, Princeton,
NJ, 1994.
13. K. Reidemeister, ‘‘Knotentheorie,’’ Chelsea, New York, 1948; Springer, Berlin, 1932;
L. F. Boron et al. ŽEnglish transl.., BSC Associates, 1983.
14. D. W. Sumners, Untangling DNA, Math. Intelligencer 12, No. 3 Ž1990., 71]80.

View publication stats