Papers
Topics
Authors
Recent
Search
2000 character limit reached

Non-Hausdorff Mapping Cylinder

Updated 21 November 2025
  • Non-Hausdorff mapping cylinder is a construction in algebraic topology and poset theory that encodes relationships between finite T₀-spaces via order relations that violate Hausdorff separation.
  • It uses the Alexandrov topology where minimal open neighborhoods act as down-sets, enabling functorial mappings and canonical inclusions that extend order-preserving maps.
  • Its homotopical properties allow the structure to collapse onto constituent spaces when inverse or forward fibers are contractible, thereby generalizing Quillen’s Theorem A and nerve theorem applications.

A non-Hausdorff mapping cylinder is a construction—originating in algebraic topology and poset theory—that generalizes the classical mapping cylinder to settings where the underlying spaces, often finite posets or Alexandroff spaces, do not satisfy Hausdorff separation axioms. The central idea is to represent the interplay between two finite T₀-spaces (or posets) linked by a relation, yielding a topological or combinatorial object that encodes both the original spaces and their interrelation while often failing to be Hausdorff. This construction is fundamental for generalizations of key homotopical results, including Quillen’s Theorem A and modern Nerve theorems, and allows a uniform combinatorial framework for topology and applied studies such as Mapper theory.

1. Construction of the Non-Hausdorff Mapping Cylinder

Given finite posets (X,X)(X, \le_X) and (Y,Y)(Y, \le_Y) and any relation RX×YR\subseteq X\times Y, the non-Hausdorff mapping cylinder (or, equivalently, the "relation cylinder" B(R)B(R)) is defined as the poset whose underlying set is the disjoint union B(R)=XYB(R)=X\sqcup Y. The ordering is as follows:

  • On elements of XX and YY, inherit their original partial orders.
  • For xXx\in X and yYy\in Y, declare xyx\leq y in B(R)B(R) whenever there exist xXxx'\geq_X x and yYyy'\leq_Y y such that (x,y)R(x',y')\in R.

This construction extends the classical mapping cylinder of an order-preserving map f:XYf:X\to Y, which is recovered when RR is the graph Γ(f)={(x,f(x))}\Gamma(f)=\{(x, f(x))\}, yielding B(f)B(f) with xyx\leq y if and only if f(x)Yyf(x)\leq_Y y in addition to the native orders on XX and YY (Fernández et al., 2018, Das et al., 2024).

The topology on B(R)B(R) is the Alexandrov topology: minimal open neighborhoods correspond to down-sets for each point. For xXx\in X, Ux={z:zx}U_x=\{z: z\leq x\}; for yYy\in Y, Uy={z:zy}U_y=\{z: z\leq y\}. In general, B(R)B(R) is not Hausdorff; separation fails when xyx\leq y is introduced via RR, forbidding disjoint open neighborhoods for xx and yy.

2. Functoriality and Universal Properties

The assignment RB(R)R\mapsto B(R) is functorial, defining a functor from the category of posets with relations (Rel(Posets)) to the category of Posets. This functor admits natural transformations—canonical inclusions iX:XB(R)i_X: X\to B(R) and jY:YB(R)j_Y:Y\to B(R). For any poset QQ and order-preserving maps α:XQ\alpha:X\to Q, β:YQ\beta:Y\to Q satisfying the compatibility xRy    α(x)Qβ(y)xRy\implies \alpha(x)\leq_Q \beta(y), there exists a unique order-preserving map Φ:B(R)Q\Phi:B(R)\to Q extending both α\alpha and β\beta. Thus, B(R)B(R) serves as the pushout in the 2-category of posets of the diagram XRYX\leftarrow R\rightarrow Y (Fernández et al., 2018).

3. Homotopical and Collapse Properties

The central homotopical feature of the non-Hausdorff mapping cylinder is its ability to interpolate and relate the homotopy types of XX and YY. For any yYy\in Y and xXx\in X, define the fibers:

  • R1(Uy)={xX:xRy for some yYy}R^{-1}(U_y)=\{x\in X: xRy'\text{ for some } y'\le_Y y\},
  • R(Fx)={yY:xRy for some xXx}R(F_x)=\{y\in Y: x'Ry\text{ for some } x'\ge_X x\}.

The following key properties hold:

  • If each R1(Uy)R^{-1}(U_y) is contractible (or collapsible), then B(R)B(R) collapse-retracts to XX and their order complexes are homotopy equivalent.
  • If each R(Fx)R(F_x) is contractible (or collapsible), then B(R)B(R) collapses onto YY.
  • If both fiber families are contractible (or collapsible), XX, B(R)B(R), and YY are all mutually simple-homotopy equivalent (Das et al., 2024).

This collapse mechanism underlies generalizations of Quillen’s Theorem A and is essential in proofs of Nerve theorems for posets and finite spaces (Fernández et al., 2018).

4. Multiple Cylinder of Relations

The concept generalizes to sequences of spaces and relations. Let X0,,XnX_0,\dots,X_n be finite T₀-spaces, with relations RiXi×Xi+1R_i\subseteq X_i\times X_{i+1}. The multiple cylinder B(R0,,Rn1;X0,,Xn)B(R_0,\dots,R_{n-1};X_0,\dots,X_n) is the union Xi\bigsqcup X_i with native ordering on each XiX_i, and comparabilities xyx\leq y (for xXix\in X_i, yXi+1y\in X_{i+1}) whenever there exist xxx'\geq x, yyy'\leq y with xRiyx' R_i y'. No additional cross-level comparabilities are introduced.

If the composite of the relations Rn1R0R_{n-1}\circ\dots\circ R_0 has all inverse fibers contractible, then the multiple cylinder collapses to X0X_0; similarly, if all forward fibers are trivial, it collapses to XnX_n. This construction allows the comparison and transfer of homotopical data across chains of spaces (essential in advanced Nerve theorem arguments and complexes arising in Mapper-type constructions) (Das et al., 2024).

5. Comparison with Classical (Hausdorff) Mapping Cylinder and Adjunction Spaces

In classical topology, the mapping cylinder M(f)=(X×[0,1])fYM(f) = (X\times[0,1])\cup_f Y (identifying (x,1)f(x)(x,1)\sim f(x)) is Hausdorff when XX and YY are, and the gluing is along closed subspaces without boundaries. However, when gluing along a region with boundary or a non-closed subspace, Hausdorffness fails precisely at those boundary points. This behavior is formalized in the adjunction-space theory: Hausdorff violations in MϕNM\cup_\phi N occur exactly at pairs of boundary points of the gluing regions (O'Connell, 2020). In the finite (combinatorial) setting, the non-Hausdorff mapping cylinder is inherently non-Hausdorff except in trivial situations. Its up-set/Alexandroff topology reflects this, and no separation axiom beyond T0T_0 typically holds.

The table below contrasts the two approaches:

Aspect Classical Mapping Cylinder Non-Hausdorff Mapping Cylinder (Relation Cylinder)
Underlying Set X×[0,1]YX\times[0,1]\cup Y XYX\sqcup Y
Topology Hausdorff (if gluing is “tame”) Alexandroff; rarely Hausdorff
Gluing Mechanism Points (x,1)f(x)(x,1)\sim f(x) Cross-relations xyx\leq y via RR-links
Homotopy Collapses Retraction onto YY always possible Collapses to XX or YY under fiber triviality

6. Applications to Homotopy Theory and Nerve Theorems

The non-Hausdorff mapping cylinder provides a framework for generalizing Quillen’s Theorem A to relations beyond order-preserving maps. Theorem 2.6 of Fernández–Minian states: if for all yYy\in Y, R1(Uy)R^{-1}(U_y) is contractible, and for all xXx\in X, R(Fx)R(F_x) is contractible, then the classifying complexes of XX and YY are simple-homotopy equivalent.

This facilitates new versions of the Nerve Theorem. Given a cover {Ui}\{U_i\} of a poset XX, construct a relation RX×(Nerve(I))opR\subset X\times (\mathrm{Nerve}(I))^{op} by xRσx R \sigma iff xiσUix\in \bigcap_{i\in \sigma} U_i. Even when intersections are not globally contractible but decompose into contractible components, the completion of the nerve (labeling each simplex with a contractible component) achieves equivalence of simple-homotopy types. These principles extend naturally to CW complexes and simplicial complexes via the associated order complexes, providing unification between classical topological theorems, Mapper-style invariants, and their combinatorial analogues (Fernández et al., 2018, Das et al., 2024).

7. Concrete Examples and Structural Features

Non-Hausdorffness is transparent in explicit constructions:

  • For X={x1<x2}X = \{x_1 < x_2\}, Y={y1<y2}Y = \{y_1 < y_2\}, R={(x1,y1),(x1,y2)}R = \{(x_1, y_1), (x_1, y_2)\}, the cylinder B(R)B(R) has x1y1x_1 \leq y_1, x1y2x_1 \leq y_2, x1x2x_1 \leq x_2, y1y2y_1 \leq y_2, and minimal open neighborhoods for y2y_2 contain both x1x_1 and y1y_1, while y1y_1 is in the closure of {x1}\{x_1\}—demonstrating inseparability (Fernández et al., 2018).
  • For the boundary of a triangle (1-skeleton of Δ2\Delta^2) and a 2-piece cover with intersections that are not contractible, the completion of the nerve (based on the mapping cylinder) restores the correct simple-homotopy type, while the classical nerve fails.

Maximal Hausdorff subspaces in non-Hausdorff mapping cylinders decompose naturally: one component from the "open cylinder" (e.g., X×[0,1)X\times[0,1)) and another from the target with the problematic glued-in boundaries removed. This decomposition is described rigorously in the adjunction space formalism (O'Connell, 2020).


References:

  • Fernández, X., Minian, E. G. "The cylinder of a relation and generalized versions of the Nerve Theorem" (Fernández et al., 2018)
  • O’Connell, J. "Non-Hausdorff Manifolds via Adjunction Spaces" (O'Connell, 2020)
  • Recent developments and multiple-relation cylinders: (Das et al., 2024)

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Non-Hausdorff Mapping Cylinder Construction.