Harmonic Archipelago in Wild Topology

• Harmonic Archipelago is a wild two-dimensional continuum formed by adjoining disks to the Hawaiian earring, showcasing non-CW complex behavior.
• It exhibits unique algebraic features, including a first singular homology isomorphic to (∏ℤ)/(⊕ℤ) and nontrivial second Čech cohomology despite vanishing singular 2-homology.
• Its fundamental group, described via archipelago groups with infinite reduced words, highlights the contrast between local complexity and global shape invariants.

Searching arXiv for the cited Harmonic Archipelago papers to ground the article in published work. The Harmonic Archipelago is a classical wild space in algebraic topology, introduced as a two-dimensional continuum obtained from the Hawaiian earring by adjoining a sequence of disks between consecutive loops, with the disks accumulating at a distinguished point. In the literature it appears both as a geometric subset of $\mathbb{R}^3$ and as an archipelago space defined via a mapping cone construction. Its significance lies in the interaction between infinite local complexity and global algebraic invariants: its singular homology, Čech cohomology, and fundamental group exhibit phenomena that depart sharply from the behavior of CW complexes and manifolds. In particular, the space is path-connected, fails to be semilocally simply connected at its accumulation point, has first singular homology $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$, and has nontrivial second Čech cohomology despite vanishing singular homology in degree $2$ (Karimov et al., 2012).

1. Definition and geometric models

The starting point is the Hawaiian earring,

$$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$

the union of a null sequence of circles tangent at the origin. It is a planar Peano continuum, $1$-dimensional, and not semilocally simply connected at the origin (Karimov et al., 2012).

Informally, the Harmonic Archipelago $\mathcal{HA}$ is obtained by adjoining a sequence of “tall disks” between consecutive loops of $\mathbb{H}$. For homotopy-theoretic calculations, one replaces $\mathcal{HA}$ by a homotopy equivalent subspace $HA\subset\mathbb{R}^3$, called the formal Harmonic Archipelago, given explicitly by

$$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$

together with cones on these circles with vertices at height $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$0: $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$1 where

$\prod \mathbb{Z}/\bigoplus \mathbb{Z}$2

and $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$3 (Karimov et al., 2012). The intervals along the $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$4-axis supply bridges between successive cones and arrange the “archipelago” structure.

A second, equivalent topological description is given by the classical archipelago model: a non-contractible subset of $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$5 homeomorphic to a disk except at one non-manifold point, built from the Hawaiian earring by attaching disks that bulge to constant height (Conner et al., 2014). The same paper shows that the harmonic archipelago is homeomorphic to the reduced metric suspension of the topologist’s sine curve and is homotopy equivalent to a mapping cone $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$6 of the canonical map from the ordinary wedge $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$7 to the shrinking wedge $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$8 (Conner et al., 2014).

These models are used differently. The explicit $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$9 is adapted to Mayer–Vietoris and Čech-cohomological calculations, whereas the mapping-cone model is adapted to the computation of $2$0 as a quotient of the Hawaiian earring group (Karimov et al., 2012, Conner et al., 2014).

2. Local topology and wild behavior

The Harmonic Archipelago is path-connected, but its local structure at the accumulation point is singular. Like the Hawaiian earring, it is not semilocally simply connected at the distinguished point $2$1: arbitrarily small neighborhoods of $2$2 contain infinitely many essential loops (Karimov et al., 2012). In the Bogley–Sieradski model, all wildness is concentrated at a distinguished point $2$3 on the boundary; moreover, $2$4 is homeomorphic to a disk with a boundary point removed (Corson, 15 Sep 2025).

This concentration of wildness has several consequences. Standard CW-complex techniques are not directly applicable, because the space is formed from a nonlocally finite family of $2$5-cells accumulating at one point and is not a CW complex (Karimov et al., 2012). At the same time, loops can be pushed into arbitrarily small neighborhoods of the singular point, while some global loops remain non-null-homotopic. In the Bogley–Sieradski description, one may homotope every loop into arbitrarily small neighborhoods of $2$6, yet a loop running around the outer boundary and encircling all hills is not null-homotopic (Corson, 15 Sep 2025).

A common misconception is that the attached $2$7-cells trivialize all higher-dimensional structure. The singular-homological calculation indeed gives $2$8 for $2$9, but the Čech cohomology computation shows that $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$0 is nonzero (Karimov et al., 2012). This indicates that the space retains shape-theoretic $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$1-dimensional information not visible to singular homology. A plausible implication is that the archipelago’s “two-dimensionality” is concentrated in the inverse-limit behavior of its neighborhoods rather than in ordinary singular $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$2-cycles.

3. Singular homology

The singular homology of the Harmonic Archipelago is

$$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$3

(Karimov et al., 2012). The identification of $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$4 reflects path-connectedness. The computation of $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$5 is the central singular-homological result.

The proof uses a Mayer–Vietoris decomposition

$$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$6

The subset $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$7 deformation retracts onto a modified Hawaiian earring $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$8 and hence is homotopy equivalent to $$\mathbb{H}=\bigcup_{n\in\mathbb{N}}\left\{(x,y)\in\mathbb{R}^2\mid x^2+\Bigl(y-\tfrac{1}{n}\Bigr)^2=\Bigl(\tfrac{1}{n}\Bigr)^2\right\},$$9. The subset $1$0 is homotopy equivalent to a countable discrete union of points, so $1$1 and $1$2. The intersection $1$3 is homotopy equivalent to a countable disjoint union of circles, with

$1$4

(Karimov et al., 2012).

The relevant fragment of the Mayer–Vietoris sequence is

$1$5

Higher singular homology vanishes because

$1$6

where $1$7 ranges over Peano subcontinua, and there is a cofinal sequence of such subcontinua each homotopy equivalent to $1$8. By Curtis–Fort for one-dimensional spaces, $1$9 for $\mathcal{HA}$0, hence $\mathcal{HA}$1 for $\mathcal{HA}$2 (Karimov et al., 2012).

The decisive algebraic input is the split exact sequence of Eda–Kawamura for the Hawaiian earring: $\mathcal{HA}$3 Here $\mathcal{HA}$4 records winding numbers around the constituent circles. Combining this with the Mayer–Vietoris map and a splitting argument via algebraic compactness yields a short exact sequence

$\mathcal{HA}$5

which splits; since $\mathcal{HA}$6 is isomorphic to its direct sum with itself, one obtains

$\mathcal{HA}$7

(Karimov et al., 2012).

This coincides with the abelianization computed in later work from the fundamental-group perspective: the abelianization of the archipelago group is

$\mathcal{HA}$8

(Corson, 15 Sep 2025).

4. Čech cohomology and inverse-limit structure

The Čech cohomology groups of the Harmonic Archipelago are

$\mathcal{HA}$9

(Karimov et al., 2012). The nontrivial $\mathbb{H}$0 is the key feature distinguishing the archipelago from several related wild spaces.

The calculation proceeds by an exhaustion $\mathbb{H}$1 by open sets, with

$\mathbb{H}$2

from the Mayer–Vietoris decomposition and, for $\mathbb{H}$3,

$\mathbb{H}$4

Each $\mathbb{H}$5 deformation retracts onto a modified Hawaiian earring, hence $\mathbb{H}$6 (Karimov et al., 2012). Therefore

$\mathbb{H}$7

The derived-limit exact sequence for a countable increasing open cover,

$\mathbb{H}$8

then becomes the basic computational tool (Karimov et al., 2012). For $\mathbb{H}$9 one gets

$\mathcal{HA}$0

The bonding maps on $\mathcal{HA}$1 are right shifts,

$\mathcal{HA}$2

For the inverse system

$\mathcal{HA}$3

one has $\mathcal{HA}$4, while

$\mathcal{HA}$5

hence

$\mathcal{HA}$6

(Karimov et al., 2012). Similarly, applying the same exact sequence with $\mathcal{HA}$7 gives $\mathcal{HA}$8, and since $\mathcal{HA}$9, all higher Čech groups vanish.

This mechanism is shape-theoretic rather than singular. The nontrivial $HA\subset\mathbb{R}^3$0 term records the accumulation of the $HA\subset\mathbb{R}^3$1-dimensional “bridging” structure through an inverse system of $HA\subset\mathbb{R}^3$2-dimensional approximants. This suggests that the Harmonic Archipelago is a paradigmatic example where derived inverse limits detect information invisible to ordinary singular homology.

5. Fundamental group and archipelago groups

The fundamental group of the harmonic archipelago is most naturally described through the theory of archipelago groups. If $HA\subset\mathbb{R}^3$3 are pointed spaces with $HA\subset\mathbb{R}^3$4, then the archipelago space is defined as the mapping cone of the canonical map

$HA\subset\mathbb{R}^3$5

from the ordinary wedge to the shrinking wedge (Conner et al., 2014). Theorem 5 of that work gives

$HA\subset\mathbb{R}^3$6

where $HA\subset\mathbb{R}^3$7 is the topologist’s product and $HA\subset\mathbb{R}^3$8 is the free product. For $HA\subset\mathbb{R}^3$9, this yields the classical archipelago group

$$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$0

which is the fundamental group of the harmonic archipelago up to homotopy (Conner et al., 2014).

The topologist’s product can be represented by infinite reduced words in which each factor appears only finitely often. In this language, the quotient by the free product “kills” all finite words and retains genuinely infinite reduced words (Conner et al., 2014). Geometrically, attaching the bulging disks makes finite concatenations null-homotopic while preserving infinite limiting behavior.

A later combinatorial formulation expresses the same group as

$$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$1

where $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$2 denotes reduced infinite words in the alphabet $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$3 and $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$4 consists of words using only one generator $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$5 (Corson, 15 Sep 2025). In this description, loops around a single island may be deleted, but infinite concatenations across indices remain.

Several structural properties of this group are established in the archipelago-group literature. For $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$6, one has:

• $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$7 embeds in $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$8 (Conner et al., 2014).
• $$C_n=\left\{(x,y,0)\in\mathbb{R}^3\mid x^2+\Bigl(y-\frac{1}{n}\Bigr)^2=\Bigl(\frac{1}{3n(n+1)}\Bigr)^2\right\}\qquad(n\in\mathbb{N}),$$9 is not indicable; equivalently, $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$00 (Conner et al., 2014).
• $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$01 is locally free, hence torsion-free (Conner et al., 2014).

The later paper on the double cone group records the same isomorphism class as having cardinality $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$02, being locally free and torsion-free, containing every countable locally free group as a subgroup, being neither free nor finitely generated, and having automorphism group containing a copy of $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$03 (Corson, 15 Sep 2025). It also states that every homomorphism from this group to $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$04 is trivial, in agreement with the non-indicability result (Corson, 15 Sep 2025, Conner et al., 2014).

6. Comparison with related spaces

The Harmonic Archipelago is closely related to both the Hawaiian earring and the Griffiths space, but it is not homotopy equivalent to the latter. Eda had shown that

$\prod \mathbb{Z}/\bigoplus \mathbb{Z}$05

so singular homology alone does not distinguish them (Karimov et al., 2012). However, the Griffiths space $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$06 is cell-like and therefore has trivial shape, with

$\prod \mathbb{Z}/\bigoplus \mathbb{Z}$07

whereas

$\prod \mathbb{Z}/\bigoplus \mathbb{Z}$08

Since Čech cohomology is a homotopy invariant in this setting, $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$09 and $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$10 are not homotopy equivalent (Karimov et al., 2012).

At the level of fundamental groups, however, the relation is much closer. The 2025 result of Corson proves Cannon–Conner’s conjecture that the fundamental group of the Griffiths double cone space is isomorphic to that of the harmonic archipelago: $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$11 (Corson, 15 Sep 2025). The proof is combinatorial and non-constructive; the paper explicitly states that there is no continuous map between $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$12 and $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$13 inducing this isomorphism on $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$14 (Corson, 15 Sep 2025).

This combination of facts is central to the space’s role in wild topology. The Harmonic Archipelago and Griffiths space have isomorphic first singular homology and isomorphic fundamental groups, yet are not homotopy equivalent because Čech cohomology distinguishes them (Karimov et al., 2012, Corson, 15 Sep 2025). This is a particularly sharp illustration of the gap between group-theoretic and shape-theoretic invariants in non-CW settings.

7. Broader significance and classification phenomena

The Harmonic Archipelago functions as the prototype of an extensive algebraic class of archipelago groups. For a countable family of groups $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$15, the archipelago group is defined by

$\prod \mathbb{Z}/\bigoplus \mathbb{Z}$16

and for countable nontrivial groups a striking classification holds: if only finitely many $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$17 have elements of order $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$18, then $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$19; if infinitely many have elements of order $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$20, then $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$21 (Conner et al., 2014). The same paper conjectures that $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$22 (Conner et al., 2014).

Corson’s later theorem enlarges this picture. For any cardinal $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$23 with $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$24, the groups $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$25 are all isomorphic, and for any sequence $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$26 of groups without involutions satisfying $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$27, all archipelago groups $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$28 are mutually isomorphic (Corson, 15 Sep 2025). In particular, $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$29, $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$30, and this whole family of archipelago groups lie in the same isomorphism class (Corson, 15 Sep 2025).

Several open or explicitly unresolved issues remain. The no-involutions hypothesis in the general isomorphism theorem is stated to remain open for arbitrary sequences with possible involutions (Corson, 15 Sep 2025). The 2025 proof of $\prod \mathbb{Z}/\bigoplus \mathbb{Z}$31 is non-constructive and does not arise from a geometric map (Corson, 15 Sep 2025). Residual finiteness, Hopficity, and co-Hopfian properties are not claimed there and are described as open or subtle (Corson, 15 Sep 2025).

The overall significance of the Harmonic Archipelago is therefore twofold. Topologically, it is a canonical wild continuum in which singular homology, Čech cohomology, and homotopy type separate in nonclassical ways (Karimov et al., 2012). Algebraically, its fundamental group is the prototype archipelago group, a locally free but nonfree and non-indicable quotient of the Hawaiian earring group that sits inside a large, flexible isomorphism class (Conner et al., 2014, Corson, 15 Sep 2025).