Published on

Geometric Topology 7

Authors

Geometric Topology 7

The Kauffman Bracket

Recall that a regular isotopy between link diagrams D,DD,D' is defined by a sequence of moves R0,R1,R2,R3R_0,R_1,R_2,R_3. Let OO denote a diagram of the unknot with no crossings (a circle), and let ODOD denote the disjoint union of plane diagrams.

Proposition: The rules:

{(i) O=1,(ii) OD=(A2+A2)D,(iii) =A+A1)(\begin{cases} \text{(i)} \ \boxed{O} = 1,\\ \text{(ii)} \ \boxed{OD} = -(A^2 + A^{-2}) \boxed{D}, \\ \text{(iii)} \ \boxed{_{-}{\diagdown}^{-}} = A \boxed{\asymp} + A^{-1} \boxed{) (} \end{cases}

allow one to associate to each link diagram DD a Laurent polynomial D\boxed{D} in AA (so one can have positive and negative powers of AA and constants) that depends only on the regular isotopy class of DD, so that DD    D=DD \approx D' \implies \boxed{D} = \boxed{D'}.

The usual notation for the Kauffman bracket is D\langle D \rangle, but using boxes will make it easier to surround parts of the diagrams. In either case, it is customary not to include the variable, which is always denoted 'AA'.

Example: The bracket of the disjoint union of n1n \geq 1 circles equals (A2A2)n1(-A^2 - A^{-2})^{n-1}.

A Word About Splittings

Each boxed term in rule (iii) denotes the Kauffman bracket of an entire diagram with the strands joined inside the boxes as shown, and no changes outside the boxes. The rule relates the bracket of 33 diagrams, in which a given crossing XX is eliminated ('split') in its two possible planar ways.

If we regard the overpass as the xx-axis (in either direction), a positive splitting (whose resulting diagram has a coefficient A+1A^{+1}) corresponds to opening up a channel between the odd quadrants. It is therefore achieved by placing a 'splitting marker' between Q1Q_1 and Q3Q_3, which can be imagined to be the line with Cartesian equation y=+xy = +x. Similarly a negative splitting corresponds to joining the even quadrants so as to insert the splitting marker y=xy = -x.

Note that orientation plays no role in the definition of the bracket. In particular, the crossing XX in (iii) has no sign associated to it, and by rotating the page any crossing can be viewed as if it looks like XX. However if the underlying diagram is chess-boarded then XX acquires a chess sign, and joining the black regions will result in a splitting whose sign equals that chess sign (positive     \iff Q1,Q3Q_1,Q_3 are black).

Invariance Under R2R_2

If a diagram has more than one crossing, the order in which one applies (iii) is irrelevant. For example, one can apply (iii) to each of 22 crossings of a diagram DD to decompose D\boxed{D} into the sum of 44 terms, with coefficients:

A2, AA1=1, A1A=1 A2A^2, \ AA^{-1} = 1, \ A^{-1}A = 1 \ A^{-2}

Resolving the next 22 crossings gives 44 terms, but (ii) causes 33 of these to cancel out. Hence, the double underpass can be removed without changing the bracket polynomial.

Invariance Under R3R_3

Apply (iii) to the diagonal crossing, then R2R_2 invariance to the diagram with coefficient AA. This gives a symmetrical configuration, which establishes R3R_3 invariance.

Starting with a diagram with cc crossings, we can keep 'splitting' so as to end up with 2c2^c diagrams, each one DiD_i (the result of a state) with no crossings. Thus DiniOD_i \approx n_i O and Di=(A2A2)ni1\boxed{D_i} = (-A^2 -A^{-2})^{n_i - 1} where each nin_i where 1i2c1 \leq i \leq 2^c is a positive.

The Effect of R1R_1

Let DD' be a link diagram incorporating a twist, and let DD be the diagram in which R1R_1 has been used to eliminate the twist (keeping the ends fixed). The latter is negative in the sense its writhe-sign is 1-1, chosen to conform with our diagram of (iii). We want to relate D\boxed{D'} to D\boxed{D}.

If the crossing XX is split positively (so a splitting marker is placed horizontally), we get a single connected arc inside the box, and so DD altogether. If XX is split negatively (vertically splitting marker), we get an extra circle resulting in a diagram DODD'' \sim OD. So:

D=AD+A1D=AD+A1(A2A2)D=A3D.\boxed{D'} = A\boxed{D} + A^{-1}\boxed{D''} = A\boxed{D} + A^{-1}(-A^2-A^{-2})\boxed{D} = -A^{-3}\boxed{D}.

Now w(D)=w(D)1w(D') = w(D) - 1, so it follows that:

(A)3w(D)D=(A)3w(D)D.(-A)^{-3w(D')}\boxed{D'} = (-A)^{-3w(D)}\boxed{D}.

Conclusion: Given a diagram DD of a link LL, the quantity (A)3w(D)D(-A)^{-3w(D)}\boxed{D} is invariant by moves R0,R1,R2,R3R_0,R_1,R_2,R_3. It therefore depends only on the ambient isotopy class of LL: it is (up to change of variable) the Jones polynomial of LL.

Examples: Hopf Link and Trefoil Diagrams

Splitting the two crossings of a diagram of the Hopf link gives 222^2 states and:

Hopf=A2+2O+A2OO=(A2+A2)(A2A2)+2=A4A4.\boxed{Hopf} = A^2\boxed{\circledcirc} + 2\boxed{O} + A^{-2}{OO} = (A^2 + A^{-2})(-A^2-A^{-2}) + 2 = -A^4 - A^{-4}.

This allows us to determine the bracket of a diagram of the right-handed trefoil R31R3_1:

D=AHopf+A1unknot with two negative twists=A(A4A4)+A1(A3)2=A5A3+A7.\boxed{D} = A\boxed{\text{Hopf}} + A^{-1}\boxed{\text{unknot with two negative twists}}\\ = A(-A^4 - A^{-4}) + A^{-1}(-A^{-3})^2 = -A^5 - A^{-3} + A^{-7}.

Exercises:

  1. The bracket of the 33-leaved diagram of L31L3_1 is A7A3A5A^7 - A^3 - A^{-5}.
  2. If mDmD is the diagram DD in which all crossings have been reversed then mD\boxed{mD} is obtained from D\boxed{D} by replacing AA with A1A^{-1}.

The Jones Polynomial

We have already defined this implicitly for the diagram DD of an oriented link LL, by combining the writhe and Kauffman bracket, so the result is unchanged by all RR-moves.

The Jones polynomial V(L)V(L) (sometimes denoted VL(t)V_L(t)) is in fact a Laurent polynomial in the variable t12t^{\frac{1}{2}} that is substituted in place fo $A^-2. It was introduced by Vaughan Jones in 1984, but it was quickly understood that it can be derived from Kauffman's bracket.

Definition: If DD is a diagram for LL then V(L)=(A)3w(D)DV(L) = (-A)^{-3w(D)}\boxed{D} with A=t14A=t^{-\frac{1}{4}}.

V(L)V(L) is an invariant of the isotopy class of LL as an oriented link, but orientation is irrelevant if LL is a knot.

We will see that if LL has an odd number of components then V(t)V(t) is actually a Laurent polynomial in tt. Otherwise it will be t12t^{\frac{1}{2}} times such a Laurent polynomial. It has integer coefficients.

Exercise: The Jones polynomial of the mirror image of a link is found by subsituting t1t^{-1} in place of tt.

Examples of Knot Polynomials

Unlike the bracket, VV is a true knot invariant, so its value can be computed using any valid diagram, which is very convenient. Previous formulae imply that VV of any unknot equals 11, and:

V(L31)=(A)+9DL=A9(A7A3A5)=t4+t3+t1,V(R31)=(A)9DR=A9(A7A3A5)=t4+t3+tV(L3_1) = (-A)^{+9}\boxed{D_L} = -A^9(A^7 - A^3 -A^{-5}) = -t^{-4} + t^{-3} + t^{-1},\\ V(R3_1) = (-A)^{-9}\boxed{D_R} = -A^{-9}(A^{-7} - A^{-3} - A^5) = -t^4 + t^3 + t

in accordance with the previous exercise.

Exercise: An easier-to-memorize formula reflects the fact that 414_1 is achiral:

V(41)=t2t1+1t+t2.V(4_1) = t^{-2} - t^{-1} + 1 - t + t^2.

Results to come:

  1. If LL is a link with 1\ell \geq 1 components then VL(1)=(2)1V_L(1) = (-2)^{\ell -1}, so a knot KK has VK(1)=1V_K(1) = 1. This is easily proved using the skein relation to come.
  2. If a knot KK has a (connected, reduced) alternating diagram with cc crossings then the difference between the highest and lowest powers of tt (the 'Laurent span') in V(K)V(K) equals cc.

Examples of Link Polynomials

We can show that:

  • V(right Hopf link)=t1/2t5/2V(\text{right Hopf link}) = -t^{1/2} - t^{5/2}
  • V(Solomon link)=t9/2t5/2+t3/2t1/2V(\text{Solomon link}) = -t^{9/2} - t^{-5/2} + t^{-3/2} - t^{-1/2}
  • V(Whitehead link)=t3/2+t1/22t1/2+t3/22t5/2+t7/2V(\text{Whitehead link}) = -t^{-3/2} + t^{-1/2} - 2t^{1/2} + t^{3/2} - 2t^{5/2} + t^{7/2}
  • V(Borromean rings)=t3+3t22t1+42t+3t2t3V(\text{Borromean rings}) = -t^{-3} + 3t^{-2} - 2t^{-1} + 4 - 2t + 3t^2 - t^3

The Skein Relation

Surround a crossing of an oriented link diagram DD by a box and orient the page so that arrows enter on the left and leave on the right. There are exactly 33 ways to arrange the strands in the box with at most one crossing:

  1. Top strand entering from the left is the overpass (and bottom strand entering from the left is the underpass) (L+L_+)
  2. Top strand entering from the left is the underpass (and bottom strand entering from the left is the overpass) (LL_{-})
  3. Top strand entering from the left and bottom strand entering from the left can be pulled apart and do not share a crossing (L0L_0)

In particular there is only one way to split an oriented crossing. There is a well-known relation between the Jones polynomials of the links one can form by rearranging the strands in such a box:

Proposition: Set z=t1/2t1/2z = t^{1/2}-t^{-1/2}. Then:

t1V(L+)tV(L)=zV(L0)t^{-1}V(L_+) - tV(L_-) = zV(L_0)

A Coherent Proof

Exercise: If L±L_{\pm} are knots then L0L_0 is a link with two components. This helps explain the appearance of half-integral powers of tt as coefficients of V(L0)V(L_0).

The skein relation is proved by splitting the two crossings using Kauffman's rule (iii).

Let w0=w(L0)w_0 = w(L_0) denote the writhe of the diagram representing L0L_0. The writhe sign of the featured crossing in L±L_{\pm} is ±1\pm 1, so the writhe of the entire diagram representing L±L_{\pm} equals w0±1w_0 \pm 1. We know that:

A4L+=A5+A3)(,A4L=A3)(+A5A^4 \boxed{L_+} = A^5{\boxed{\asymp}} + A^3\boxed{)(},\\ A^{-4}\boxed{L_-} = A^{-3}\boxed{)(} + A^{-5}\boxed{\asymp}

Therefore:

t1V(L+)=(A)3w(L+)A4L+=(A)3w0(A2)()tV(L)=(A)3w(L+)A4L=(A)3w0()(A2)t^{-1}V(L_+) = (-A)^{-3w(L_+)}A^4\boxed{L_+} = (-A)^{-3w_0}(-A^2\boxed{\asymp}-\boxed{)(})\\ tV(L_-) = (-A)^{-3w(L_+)}A^{-4}\boxed{L_-} = (-A)^{-3w_0}(-\boxed{)(} - A^{-2}\boxed{\asymp})

Subtracting:

t1V(L+)tV(L)=(A2+A2)(A)3w0=zV(L0).t^{-1}V(L_+) - tV(L_-) = (-A^2 + A^{-2})(-A)^{-3w_0}\boxed{\asymp} = zV(L_0).

We have used the change of variable t=A4t = A^{-4} freely. \blacksquare

Example: The Jones Polynomial of 515_1

In order to compute the Jones polynomial of the knot 515_1, we first note it has writhe +5+5, and its crossings are of the type seen in L+L_+. Set z=t1/2tt/2z = t^{1/2} - t^{-t/2} again. If we focus on the top crossing of the knot and apply R2R_2 to obtain the version of the diagram with the top crossing flipped, we get the equation:

V(51)=t2V(R31)+tzV(S)V(5_1) = t^2V(R3_1) + tzV(S)

where SS is the right-handed Solomon link, with the right-handed orientation (w=+4w = +4).

Example: The Jones Polynomial of the Solomon Link

We can express V(S)V(S) in terms of the polynomials:

V(RH)=t5/2t1/2,V(R31)=t4+t3+tV(RH) = -t^{5/2} - t^{1/2}, \\ V(R3_1) = -t^4 + t^3 + t

of the right-handed trefoil knot and Hopf link. Focusing on the top-left crossing we get:

V(S)=t2V(RH)+tzV(R31)=t11/2+t9/2t7/2t3/2.V(S) = t^2V(RH) + tzV(R3_1) = -t^{11/2} + t^{9/2} - t^{7/2} - t^{3/2}.

Exercise: Why is this t6t^6 times the previous polynomial for the Solomon link?

Putting it all together, V(51)=t7+t6t5+t4+t2V(5_1) = -t^7 + t^6 - t^5 + t^4 + t^2.