nLab identity among the relations


We consider a presentation, 𝒫=(X:R)\mathcal{P} = (X : R), of a group GG. We thus have a short exact sequence,

1→N→F→G→1,1\to N \to F \to G \to 1,

where F=F(X)F = F(X), the free group on the set XX, RR is a subset of FF and N=N(R)N = N(R) is the normal closure in FF of the set RR. The group, FF, acts on NN by conjugation: uc=ucu βˆ’1{}^u c = u c u^{-1}, c∈Nc\in N, u∈Fu \in F, and the elements of NN are words in the conjugates of the elements of RR:

c= u 1(r 1 Ξ΅ 1) u 2(r 2 Ξ΅ 2)… u n(r n Ξ΅ n),c = {}^{u_1}(r_1^{\varepsilon_1}){}^{u_2}(r_2^{\varepsilon_2})\ldots {}^{u_n}(r_n^{\varepsilon_n}),

often says such elements are consequences of RR. Heuristically, an identity among the relations of 𝒫\mathcal{P} is such an element, cc, which equals 1.

The problem of what this actually means is analogous to that of working with a relation, since, for example, in the presentation, (a:a 3)( a : a^3), of C 3C_3, the cyclic group of order 3, if aa is thought of as being an element of C 3C_3, then a 3=1a^3 = 1! Doesn’t that say the presentation is (a:1)( a : 1)? Why is this then different from the situation with the β€˜presentation’, (a:a=1)( a : a = 1)?

To get around that mess, the free group on the generators, F(X)F(X), was introduced and, of course, in F({a})F(\{a\}), a 3a^3 is not 1. The relation is thus thought of as being a piece of data that is giving the instruction to rewrite that element to 1.

An analogous free algebraic device, namely a free crossed module on the presentation can be introduced to handle the identities.

Before giving the formal definition we will look at some examples.

Example (Proper powers)

Suppose r∈Rr\in R, but it is a power of some element s∈Fs\in F, i.e., r=s mr = s^m. Of course, rs=sr r s = s r and

srr βˆ’1=1{}^s r r^{-1} = 1

so sr.r βˆ’1{}^s r . r^{-1} is an identity. In fact, there will be a unique z∈Fz\in F with r=z qr = z^q, qq maximal with this property. This zz is called the root of rr and if q>1q \gt 1, rr is called a proper power.

Example (Standard presentation of S 3S_3)

Here we take up the example from Cayley graph:

Consider one of the standard presentations of S 3S_3,

(a,b:a 3,b 2,(ab) 2). (a,b : a^3, b^2, (ab)^2).

Write r=a 3 r = a^3, s=b 2s = b^2, t=(ab) 2t = (ab)^2. Here the presentation leads to F=F(a,b)F= F(a,b), free of rank 2, and N(R)βŠ‚FN(R) \subset F, so it must be free as well, by the Nielsen-Schreier theorem. Its rank will be 7, given by the Schreier index formula or, geometrically, it will be the fundamental group of the Cayley quiver, also called the Cayley graph, of the presentation. This group is free on generators corresponding to edges outside a maximal tree.

The set of normal generators of N(R)N(R) has 3 elements; N(R)N(R) is free on 7 elements (corresponding to the edges not in the tree), yet it is specified as consisting of products of conjugates of rr, ss and tt, and there are infinitely many of these conjugates. Clearly there must be some slight redundancy, so there must be some identities among the relations!

Note that the Cayley graph is planar.

A path around the first triangle, 1β†’aβ†’a 21 \to a \to a^2, corresponds to the relation rr; each other region corresponds to a conjugate of one of rr, ss or tt. (It may help in what follows to think of the graph being embedded on a 2-sphere, so β€˜outer’ and β€˜outside’ mean β€˜round the back face’.) Consider a loop around a region.

Pick a path to a start vertex of the loop, a path starting at 1. For instance the path that leaves 1 and goes along edges labelled aa, bb and then goes around aaaa a a before returning by b βˆ’1a βˆ’1b^{-1} a^{-1} gives abrb βˆ’1a βˆ’1a b r b^{-1} a^{-1}. Now the path around the outside can be written as a product of paths around the inner parts of the graph, e.g. (abab)b βˆ’1a βˆ’1b βˆ’1(bb)(b βˆ’1a βˆ’1b βˆ’1a βˆ’1)…(a b a b) b^{-1} a^{-1} b^{-1} (b b) (b^{-1}a^{-1} b^{-1} a^{-1}) \ldots and so on, thus rr can be written in a non-trivial way as a product of conjugates of rr, ss and tt. (An explicit identity constructed like this is given in the paper by Brown and Huebschmann (see below).)

Example (Free Abelian group β„€ 3\mathbb{Z}^3)

In a presentation of the free Abelian group on 3 generators, one would expect the commutators, [x,y][x,y], [x,z] [x,z] and [y,z][y,z]. There is a well-known identity, that expands out to give an identity among these relations (again see Brown-Huebschman (below), p.154, or Loday, (again below) for a complete treatment.)

In a bit more detail, look at this as a presented group,

G=β„€ 3=⟨x,y,z∣[y,z],[z,x],[x,y]⟩G=\mathbb{Z}^{3}=\langle x,y,z\mid [y,z],[z,x],[x,y]\rangle

then label the relators r x=[y,z]r_x= [y,z], r y=[z,x]r_y=[z,x], r z=[x,y]r_z= [x,y].

There is an identity:

[[x,y], yz].[[y,z], zx].[[z,x], xy]=1[[x,y],{}^y z].[[y,z],{}^z x].[[z,x],{}^x y] =1

which is easily checked to hold just by expanding it out and cancelling terms. This is one of the (many) forms of the well-known identity known variously as the Hall-Witt identity, the Jacobi-Hall-Witt identity, the Jacobi identity, and various variants of these. There is thus a word,

[r z, yz].[r x, zx].[r y, xy][r_z,{}^y z].[r_x,{}^z x].[r_y,{}^x y]

which can be written out as a product of conjugates of the three relations, r xr_x, r yr_y and r zr_z, and which evaluates to 1 when considered as an element of F(x,y,z)F(x,y,z). It is thus another case of an identity among the relations for this presented group.

Note the Cayley quiver of this presentation is infinite.


We can now formally define the module of identities of a presentation, 𝒫=(X:R)\mathcal{P} = (X : R).

We first form the free crossed module on the mapping Rβ†’F(X)R\to F(X), which we will denote by βˆ‚:C(𝒫)β†’F(X)\partial : C(\mathcal{P}) \to F(X).


The module of identities , ΞΊ(𝒫)\kappa(\mathcal{P}), of 𝒫\mathcal{P} is Kerβˆ‚Ker\, \partial.

An identity among the relations of 𝒫\mathcal{P} is then an element of ΞΊ(𝒫)\kappa(\mathcal{P}).

By construction, the group presented by 𝒫\mathcal{P} is Gβ‰…F(X)/Imβˆ‚G \cong F(X)/Im\, \partial, where Imβˆ‚Im \,\partial is just the normal closure of the set, RR, of relations and we know that Kerβˆ‚Ker\, \partial is a GG-module. In fact we have an exact sequence:

0β†’ΞΊ(𝒫)β†’C(𝒫)β†’βˆ‚F(X)β†’Gβ†’1.0\to \kappa(\mathcal{P}) \to C(\mathcal{P}) \xrightarrow{\partial} F(X)\to G\to 1.


  • R. Brown and J. Huebschmann, Identities among relations, in R.Brown and T.L.Thickstun, eds., Low Dimensional Topology, London Math. Soc Lecture Notes, Cambridge University Press, (1982).

  • M. Kapranov and M. Saito, 1999, Hidden Stasheff polytopes in algebraic K-theory and in the space of Morse functions, in Higher homotopy structure in topology and mathematical physics (Poughkeepsie, N.Y. 1996), volume 227 of Contemporary Mathematics, 191 – 225, AMS

  • J.-L. Loday, 2000, Homotopical Syzygies, in Une dΓ©gustation topologique: Homotopy theory in the Swiss Alps, volume 265 of Contemporary Mathematics, 99 – 127, AMS.

Last revised on October 12, 2019 at 15:42:21. See the history of this page for a list of all contributions to it.