---
title: Topological Join Operation
url: https://www.emergentmind.com/topics/topological-join-operation
type: topic
---

# Topological Join Operation

Searching arXiv for recent and foundational papers on topological joins and related formulations.
Using the arXiv search tool to retrieve papers on topological joins, geometric joins, and join theorems.
The topological join operation is a construction that combines spaces, simplicial complexes, or related objects by adjoining a new interval parameter and collapsing opposite ends onto the two factors. In its classical form, for spaces \(A\) and \(B\), the join is
\[
A*B=\frac{A\times B\times[0,1]}{\{(a,b,0)\sim(a,b',0)\}\cup\{(a,b,1)\sim(a',b,1)\}},
\]
while for finite simplicial complexes with disjoint vertex-sets it is the abstract join
\[
K*L=\{\sigma\cup\tau\mid \sigma\in K,\ \tau\in L\}.
\]
Across algebraic topology, topological combinatorics, TDA, and singularity theory, the join serves as a dimension-raising operation with a characteristic homological shift, and several distinct but related formulations coexist: the simplicial join used in chain-level constructions [1506.06058], the classical quotient-space join and its geometric realizations [1309.0920], the augmental join with a strict \((-1)\)-dimensional unit [1212.6169], and join theorems describing Milnor fibres of sums of real-analytic germs [2002.06549].

## 1. Formal definitions and principal variants

For finite simplicial complexes \(K\) and \(L\) with disjoint vertex-sets, the abstract join is defined by
\[
K*L=\{\sigma\cup\tau\mid \sigma\in K,\ \tau\in L\}.
\]
If \(\sigma=[v_0,\dots,v_p]\in C_p(K)\) and \(\tau=[w_0,\dots,w_q]\in C_q(L)\), one orders the union of vertices as
\[
\sigma*\tau=[v_0,\dots,v_p,w_0,\dots,w_q]\in C_{p+q+1}(K*L),
\]
which already exhibits the characteristic degree shift by \(+1\) [1506.06058].

For topological spaces, the reduced join is the quotient
\[
A*B=\frac{A\times B\times[0,1]}{\{(a,b,0)\sim(a,b',0)\}\cup\{(a,b,1)\sim(a',b,1)\}},
\]
and it satisfies the formal property
\[
\widetilde H_*(A*B)\cong\widetilde H_*(\Sigma(A\wedge B)).
\]
This quotient-space description makes explicit that the construction “cones off” each factor at opposite ends of the interval [1309.0920].

A distinct Euclidean incarnation is the geometric join. For subsets \(X_1,\dots,X_k\subset\mathbb R^d\),
\[
X_{[k]}=\Bigl\{t_1x_1+\cdots+t_kx_k\in\mathbb R^d:x_i\in X_i,\ t_i\ge 0,\ \sum_{i=1}^k t_i=1\Bigr\},
\]
equivalently the union of all colorful simplices, that is, convex hulls of one point from each \(X_i\). One may realize the abstract join of the discrete sets \(X_i\) as a simplicial complex and linearly map it into \(\mathbb R^d\); its image is \(X_{[k]}\). In general the homotopy type of the geometric join can be strictly simpler than that of the abstract join [1309.0920].

Fors introduced an augmental reformulation in the category \(Top_\wp\), where the new isolate \(\mathbb 1_k=\{6\}\) is the unique \((-1)\)-dimensional space and serves as a strict join-unit. In that setting the augmental join \(X\odot Y\) is defined by a pushout, equivalently by
\[
X\odot Y\cong (X\times Y\times[0,1])/\sim,
\]
with the boundary identifications collapsing \(X\times Y\times\{0\}\) onto \(X\sqcup\{6\}\) and \(X\times Y\times\{1\}\) onto \(Y\sqcup\{6\}\) [1212.6169].

## 2. Chain-level structure and homological consequences

Over \(\mathbb Z/2\)-coefficients, the simplicial join induces a bilinear chain map
\[
*:C_p(K)\otimes C_q(L)\to C_{p+q+1}(K*L),\qquad \sigma\otimes\tau\mapsto \sigma*\tau.
\]
If
\[
c=\sum_{i=1}^m\sigma_i\in C_p(K),\qquad d=\sum_{j=1}^n\tau_j\in C_q(L),
\]
then
\[
c*d=\sum_{i=1}^m\sum_{j=1}^n(\sigma_i*\tau_j).
\]
In particular, if \(c\) and \(d\) are cycles then \(c*d\) is again a cycle [1506.06058].

The boundary formula over \(\mathbb Z/2\) is
\[
\partial(c*d)=(\partial c)*d+c*(\partial d).
\]
Over a general field \(F\), with ordered vertices, the signed formula becomes
\[
\partial(\sigma*\tau)=(\partial\sigma)*\tau+(-1)^{\dim \sigma+1}\sigma*(\partial\tau).
\]
The \(\mathbb Z/2\)-setting eliminates the sign bookkeeping but not the degree shift [1506.06058].

A convenient inverse “join-projection” chain map is
\[
f:C_{r+1}(K*L)\to C(K)\otimes C(L),
\]
specified by
\[
f(\sigma*\tau)=\sigma\otimes\tau,\qquad f(\sigma)=\sigma\otimes 1,\qquad f(\tau)=1\otimes\tau,\qquad f(1)=1\otimes 1,
\]
where \(1\) denotes the unique \((-1)\)-simplex in the augmented complex. It satisfies
\[
f\circ\partial=(\partial\otimes id+id\otimes\partial)\circ f.
\]
This map is central in the homological analysis of joined cycles [1506.06058].

Ellis’s Theorem 1.2 states: if \(a\in H_p(K)\) and \(b\in H_q(L)\) are nontrivial classes represented by cycles \(w\in C_p(K)\) and \(z\in C_q(L)\), and if \(w*z\) lies in a subcomplex \(M\subset K*L\) and represents a nonzero class in \(H_{p+q+1}(M)\), then the inclusion-induced map
\[
i_*:H_{p+q+1}(M)\to H_{p+q+1}(K*L)
\]
carries \([w*z]\) to a nontrivial class. The proof uses the composition
\[
C(M)\hookrightarrow C(K*L)\xrightarrow{f} C(K)\otimes C(L)
\]
and the Künneth theorem over a field, under which
\[
H_{p+q+1}(K*L)\to H_p(K)\otimes H_q(L)
\]
sends \([w*z]\mapsto [w]\otimes[z]\neq 0\). Proposition 1.3 gives the converse: if \(w*z\) is non-bounding in \(K*L\), then \(w\) and \(z\) must already represent nontrivial homology in \(K\) and \(L\), respectively [1506.06058].

## 3. Monoidal and augmental formulations

In the augmental category \(Top_\wp\), the join is not merely associative and symmetric up to homotopy; it is organized as a closed symmetric monoidal structure. There is a faithful functor
\[
F_\wp:Top\to Top_\wp,\qquad F_\wp(X)=X\sqcup\{6\},
\]
and the augmental join-unit is
\[
\mathbb 1_k=\{6\}\subset Top_\wp.
\]
The three point-free objects are explicitly distinguished: \(\emptyset\) as classical empty space and join-zero, \(\{\emptyset\}\) as simplicial join-unit in \(\Delta\)-complexes, and \(\mathbb 1_k\) as the new topological join-unit of dimension \(-1\) [1212.6169].

The operation \(\odot\) satisfies canonical associativity,
\[
(X\odot Y)\odot Z\cong X\odot(Y\odot Z),
\]
strict unit laws,
\[
X\odot\mathbb 1_k\cong X\cong \mathbb 1_k\odot X,
\]
and symmetry,
\[
X\odot Y\cong Y\odot X.
\]
Moreover, there is a natural homeomorphism
\[
Top_\wp(X\odot Y,Z)\cong Top_\wp(X,Map_\wp(Y,Z)),
\]
so \((Top_\wp,\odot,\mathbb 1_k)\) is a closed symmetric monoidal category [1212.6169].

This formulation aligns topological joins with augmented simplicial constructions. If \(\Sigma_1,\Sigma_2\) are \(\Delta\)-complexes with geometric realizations \(|\Sigma_i|\in Top_\wp\), then
\[
|\Sigma_1*\Sigma_2|\cong |\Sigma_1|\odot |\Sigma_2|,\qquad |\Sigma_1\times\Sigma_2|\cong |\Sigma_1|\times|\Sigma_2|.
\]
Fors further states that the Stanley–Reisner functor \(k[-]\) carries simplicial join to tensor product, and geometric realization carries that back to topological join [1212.6169].

Several classical identities appear naturally in this setting. For \(m,n\ge 0\),
\[
S^m\odot S^n\cong S^{m+n+1},
\]
and for disjoint manifolds \(M_1,M_2\),
\[
\partial(M_1\odot M_2)=(\partial M_1\odot M_2)\cup(M_1\odot\partial M_2).
\]
The join also admits a Künneth theorem with the characteristic degree shift:
\[
0\to \bigoplus_{i+j=q}H_i(X_1,X_2;G)\otimes H_j(Y_1,Y_2;G')\to H_{q+1}(X_1,X_2\odot Y_1,Y_2;G\otimes G')\to \bigoplus_{i+j=q-1}Tor_R(H_i(X_1,X_2;G),H_j(Y_1,Y_2;G'))\to 0,
\]
under the stated excision and torsion hypotheses [1212.6169].

## 4. Geometric joins, connectivity, and contractibility thresholds

The geometric join links join constructions to convexity and discrete geometry. It arises naturally in the colorful Carathéodory and Tverberg theorems, and in transversal-Helly-type questions. Its central conjecture, due to Bárány–Holmsen–Karasev, is that if \(k\ge d+1\), then for any subsets \(X_1,\dots,X_k\subset\mathbb R^d\), the geometric join \(X_{[k]}\) is contractible. This remains open even for \(k=d+1\) when each \(X_i\) has only two points [1309.0920].

Exact results are known in low dimensions. For \(d=2\), whenever \(k\ge 3\), the union of all colorful triangles is star-shaped, hence contractible. For \(d=3\), Theorem 6.1 states: if \(d=3\) and \(k\ge 4\), then \(X_{[k]}\) is contractible. The proof combines simple connectedness with vanishing \(H_2\) in a connected \(3\)-dimensional union of simplices [1309.0920].

In arbitrary dimension, stronger but higher-threshold sufficient conditions are known. Theorem 4.1 states that if \(k>d(d+1)\), then \(X_{[k]}\) is star-shaped, hence contractible. More generally, Theorem 4.2 states that if
\[
k>(r-1)(d+1),
\]
then \(X_{[k]}\) is \((r-2)\)-connected. The same picture extends to matroidal geometric joins:
\[
M_{r,d}=\Bigl\{\sum_{i=1}^s t_i x_i:\{x_1,\dots,x_s\}\subset E\text{ is independent in }M,\ t_i\ge 0,\ \sum_i t_i=1\Bigr\},
\]
with contractibility when \(\mathrm{rank}(M)>d(d+1)\) and simple connectedness when \(\mathrm{rank}(M)>\binom{d+1}{2}+2\) [1309.0920].

A recurring misconception is that the abstract join and the geometric join should have comparable homotopy type. The cited results explicitly warn against this: the abstract join of \(k\) non-trivial finite sets is a wedge of \((k-1)\)-spheres, whereas the geometric join \(X_{[k]}\subset\mathbb R^d\) can collapse many of those spheres and, under the conjectured threshold \(k=d+1\), would become contractible [1309.0920].

## 5. Join signatures in concurrence topology and Dowker filtrations

In concurrence topology, a TDA method for binary data, one constructs a filtration consisting of Dowker complexes and computes persistent homology. Persistent classes correspond to a form of negative statistical association among the variables. Ellis analyzes how the topological join appears when two groups of binary variables are examined separately and then combined [1506.06058].

Suppose two groups of variables display negative association individually, manifested in nontrivial concurrence homology in dimensions \(p\) and \(q\). If the two groups are statistically independent and the sample size is large, then representative cycles, one from each group, combine to produce a cycle in dimension \(p+q+1\). In the Dowker-filtration setting, at each frequency threshold one has a complex \(M\) on variables \(V_1,\dots,V_n\), with \(K\) the induced subcomplex on the first \(m\) variables and \(L\) the induced subcomplex on the last \(n-m\). If
\[
(V_1,\dots,V_m)\perp (V_{m+1},\dots,V_n)
\]
in law, then “with high probability” every simplex \(\sigma\in K\) and \(\tau\in L\) will jointly appear in \(M\), so \(M\) will look like \(K*L\). Long-lived classes in \(H_p(K)\) and \(H_q(L)\) then give rise, via the join operator, to nontrivial classes in \(H_{p+q+1}(M)\) [1506.06058].

The diagnostic proposed there compares persistence in the filtration
\[
M\to K*L
\]
and regards any class that survives at least two steps as a signature of independence. This use of joins is therefore not merely formal; it provides a topological criterion for dependence among groups of variables [1506.06058].

The limitations are explicit. One needs \(K\cap L=\emptyset\) so that every \(\sigma\cup\tau\) is a genuine simplex of \(K*L\). One also needs nontrivial factors: both \(w\in H_p(K)\) and \(z\in H_q(L)\) must be nonzero, and \(w*z\) must lie as a cycle in the test complex \(M\). Ellis’s examples further show that not every nontrivial class in a subcomplex of a join is literally of the form \(w*z\): the hexagon example gives a \(1\)-cycle that is nontrivial but is not of join form, and Example 1.4(3) gives a \(2\)-cycle nontrivial in both \(H_2(M)\) and \(H_2(K*L)\) that again is not literally a join \(w*z\) [1506.06058].

## 6. Join theorems for Milnor fibres and real-analytic singularities

Inaba’s join theorem places the topological join at the center of the topology of real-analytic singularities. Let
\[
f_j:(\mathbb R^{n_j},0)\to(\mathbb R^p,0),\qquad j=1,2,
\]
be real-analytic germs of independent variables, with \(n_j\ge p>2\), each satisfying: \(0\) is an isolated critical value and \(f_j\) satisfies the Thom-\(a_f\)-condition with respect to some Whitney stratification of \(V(f_j)=f_j^{-1}(0)\). Writing
\[
f=f_1+f_2
\]
on \(\mathbb R^{n+m}=\mathbb R^n\times\mathbb R^m\), one obtains tubular Milnor fibrations for \(f_1\), \(f_2\), and \(f\) [2002.06549].

For sufficiently small radii, the tubular Milnor fibre is
\[
F(f):=f^{-1}(t)\cap B^N_\varepsilon
\]
for any regular value \(t\in S^{p-1}_\delta\), and its homotopy type is independent of \(\varepsilon,\delta\), and of the choice of \(t\). The join theorem then states:
\[
F(f_1+f_2)\simeq F(f_1)*F(f_2).
\]
When \(p=2\), the monodromy of the fibration of \(f\) is, up to homotopy, the join of the monodromies of \(f_1\) and \(f_2\) [2002.06549].

The proof proceeds by constructing a tubular Milnor fibration for each germ, showing that \(f_1+f_2\) still satisfies \(a_f\), analyzing the projection
\[
F_1:Z_t\to\mathbb R^p,\qquad F_1(x,y)=f_1(x),
\]
over a radial segment \(J\) from \(0\) to \(t\), and then collapsing boundary pieces to obtain the mapping-cone description of the join. In particular, one finds a homeomorphism
\[
T(Z_t\cap F_1^{-1}(J))\cong X_t*Y_t,
\]
followed by a homotopy equivalence
\[
Z_t\simeq X_t*Y_t,
\]
with \(X_t\simeq F(f_1)\) and \(Y_t\simeq F(f_2)\) [2002.06549].

The theorem has several stated corollaries. For \(p=2\), if \(\widetilde\zeta_j(t)\) denotes the reduced zeta-function of the monodromy of \(f_j\), then
\[
\widetilde\zeta_{f_1+f_2}(t)=\widetilde\zeta_{f_1}(t)\cdot \widetilde\zeta_{f_2}(t).
\]
The Seifert form of the link \(K(f)\) is, up to an explicit sign, the tensor-product or “join-product” of the Seifert forms of \(K(f_1)\) and \(K(f_2)\). For Neumann–Rudolph’s enhanced Milnor invariants \((\mu,\lambda)\),
\[
\mu(K(f_1+f_2))=\mu(K(f_1))\cdot\mu(K(f_2)),
\]
and
\[
\lambda(K(f_1+f_2))\equiv \lambda(K(f_1))\cdot\mu(K(f_2))+\mu(K(f_1))\cdot\lambda(K(f_2))\mod 2.
\]
These statements place the join alongside addition of germs as a precise topological operation on Milnor fibres, monodromy, and links [2002.06549].

Source: https://www.emergentmind.com/topics/topological-join-operation