Alexander matrix 2026-09-28
Given a presentation of a knot group, apply abelianization to the Fox derivatives . The resulting matrix over is an Alexander matrix; suitable maximal minors recover the Alexander polynomial of a knot.
Orient the diagram and assign its regions an Alexander numbering with numbered zero. The abelianization sends to , where is the number of . Since is adjacent to , .
Form the square matrix
from the Fox derivatives with respect to for . This is the Alexander matrix with the column deleted. The Fox identity implies that its maximal minors differ by the factors , and the standard presentation of the Alexander module therefore gives
because . Thus is up to a unit .
For the free presentation , conjugation in gives the relation module
where is any lift of . A different lift differs by an element of , whose inner conjugation acts trivially on the abelianization, so this is a well-defined -module action.
Choose free generators of . The presentation relation sequence becomes
where . In Fox calculus, the first map is
The Fox identity gives , and the standard lifting argument in the free group proves exactness.
Split the sequence at the augmentation ideal . Applying to
and using projectivity of gives
The other short exact sequence identifies the last group with . A one-cocycle on is a derivation and is determined freely by its values on , so is modulo principal derivations. Principal derivations vanish on , and restriction sends a derivation to the -map . The preceding cokernel sequence therefore descends to the Mac Lane exact sequence for a free presentation
The left map need not be injective. Take , , , and the trivial module , where . Then , while restriction sends the derivation determined by to
Thus the left map is zero although its domain is nonzero, and is nontrivial.