ConceptComplete

Fox Calculus and the Alexander Polynomial

The Fox free differential calculus provides a systematic algebraic method for computing invariants of knot groups, most notably the Alexander polynomial. This approach replaces the geometric construction of covering spaces with pure algebra on group presentations.


Fox Derivatives

Definition6.4Fox Free Differential Calculus

Let F=F(x1,,xn)F = F(x_1,\ldots,x_n) be the free group on generators x1,,xnx_1,\ldots,x_n. The Fox derivatives xj:Z[F]Z[F]\frac{\partial}{\partial x_j}: \mathbb{Z}[F] \to \mathbb{Z}[F] are the unique Z\mathbb{Z}-linear maps satisfying: (1) xixj=δij\frac{\partial x_i}{\partial x_j} = \delta_{ij}, (2) (uv)xj=uxj+uvxj\frac{\partial(uv)}{\partial x_j} = \frac{\partial u}{\partial x_j} + u\frac{\partial v}{\partial x_j} (Leibniz rule), and (3) 1xj=0\frac{\partial 1}{\partial x_j} = 0. From (2): xi1xj=xi1δij\frac{\partial x_i^{-1}}{\partial x_j} = -x_i^{-1}\delta_{ij}. For any word w=xi1ε1xikεkw = x_{i_1}^{\varepsilon_1}\cdots x_{i_k}^{\varepsilon_k}, Fox derivatives compute the "infinitesimal effect" of each generator.

ExampleFox Derivative Computations

For w=x1x2x11w = x_1 x_2 x_1^{-1}: wx1=1+x1x2(x11)=1x1x2x11\frac{\partial w}{\partial x_1} = 1 + x_1x_2\cdot(-x_1^{-1}) = 1 - x_1x_2x_1^{-1} and wx2=x1\frac{\partial w}{\partial x_2} = x_1. For w=x13w = x_1^3: wx1=1+x1+x12\frac{\partial w}{\partial x_1} = 1 + x_1 + x_1^2. In general, xinxi=1+xi++xin1\frac{\partial x_i^n}{\partial x_i} = 1 + x_i + \cdots + x_i^{n-1} for n>0n > 0.


The Alexander Matrix and Polynomial

Definition6.5Alexander Matrix

Given a Wirtinger presentation πK=x1,,xnr1,,rn\pi_K = \langle x_1,\ldots,x_n \mid r_1,\ldots,r_n\rangle, the Alexander matrix AA is the (n1)×(n1)(n-1)\times(n-1) matrix obtained as follows: (1) Compute the Jacobian matrix Jij=rixjZ[F]J_{ij} = \frac{\partial r_i}{\partial x_j} \in \mathbb{Z}[F]. (2) Apply the abelianization map ϕ:Z[πK]Z[t±1]\phi: \mathbb{Z}[\pi_K] \to \mathbb{Z}[t^{\pm 1}] sending each generator xjtx_j \mapsto t (since πKabZ\pi_K^{\mathrm{ab}} \cong \mathbb{Z}). (3) Delete one row and one column from ϕ(J)\phi(J). The Alexander polynomial ΔK(t)\Delta_K(t) is the determinant detA\det A, defined up to units ±tk\pm t^k in Z[t±1]\mathbb{Z}[t^{\pm 1}].

ExampleAlexander Polynomial of the Trefoil

The trefoil has presentation a,baba=bab\langle a,b \mid aba = bab\rangle, equivalently a,babab1a1b1\langle a,b \mid abab^{-1}a^{-1}b^{-1}\rangle. Write r=abab1a1b1r = abab^{-1}a^{-1}b^{-1}. Fox derivatives: ra=1+ab1abab1a11=1+ababab1a1\frac{\partial r}{\partial a} = 1 + ab\cdot 1 - abab^{-1}a^{-1}\cdot 1 = 1 + ab - abab^{-1}a^{-1}. Under a,bta,b \mapsto t: 1+t2t2t2=1+t21=t21 + t^2 - t^2 \cdot t^{-2} = 1 + t^2 - 1 = t^2. Wait, let us be precise. We have r=abab1a1b1r = a \cdot b \cdot a \cdot b^{-1}\cdot a^{-1}\cdot b^{-1}. Then ra=1+ababab1a1\frac{\partial r}{\partial a} = 1 + ab - abab^{-1}a^{-1} and rb=aabab1+abab1a1b1(b1)\frac{\partial r}{\partial b} = a - aba b^{-1} + abab^{-1}a^{-1}b^{-1}\cdot(-b^{-1}). After abelianization (a,bta,b\mapsto t): the Alexander polynomial is Δ31(t)=1t+t2\Delta_{3_1}(t) = 1 - t + t^2.


Properties

Definition6.6Properties of the Alexander Polynomial

The Alexander polynomial satisfies: (1) Symmetry: ΔK(t1)=±tkΔK(t)\Delta_K(t^{-1}) = \pm t^k\Delta_K(t) for some kk (from the duality of the knot exterior). (2) Normalization: ΔK(1)=±1\Delta_K(1) = \pm 1 (since H1(S3K;Z)=ZH_1(S^3\setminus K;\mathbb{Z}) = \mathbb{Z}). (3) Skein relation: ΔK+(t)ΔK(t)=(t1/2t1/2)ΔK0(t)\Delta_{K_+}(t) - \Delta_{K_-}(t) = (t^{1/2}-t^{-1/2})\Delta_{K_0}(t) where K+K_+, KK_-, K0K_0 are links related by a crossing change. (4) Genus bound: the degree of ΔK\Delta_K satisfies degΔK2g(K)\deg\Delta_K \leq 2g(K) where g(K)g(K) is the Seifert genus. For fibered knots, equality holds and ΔK\Delta_K is monic.

RemarkLimitations and Generalizations

The Alexander polynomial cannot detect the unknot (there exist nontrivial knots with ΔK=1\Delta_K = 1, e.g., the Kinoshita-Terasaka knot 11n4211n42). It does not detect chirality (ΔK=ΔKˉ\Delta_K = \Delta_{\bar{K}}). Twisted Alexander polynomials ΔKρ(t)\Delta_K^\rho(t) associated to representations ρ:πKGL(V)\rho: \pi_K \to \mathrm{GL}(V) are strictly stronger invariants, detecting fibredness and genus in many cases. The Alexander module H1(X~K;Z)H_1(\tilde{X}_K;\mathbb{Z}) as a Z[t±1]\mathbb{Z}[t^{\pm 1}]-module carries more information than ΔK\Delta_K alone, with the polynomial being just the order of this module.