---
title: Cactus Flower Moduli Spaces
url: https://www.emergentmind.com/topics/cactus-flower-moduli-spaces
type: topic
---

# Cactus Flower Moduli Spaces

Searching arXiv for recent and foundational papers on cactus flower moduli spaces, pure cactus groups, weighted cactus groups, and related Gaudin-model connections.
I’ll verify the relevant arXiv records and publication metadata for the core papers before writing the article.
Using the arXiv API to confirm the cited papers and their metadata.
```python
import requests, textwrap, re, xml.etree.ElementTree as ET, json, pandas as pd
ids = ["1501.02835","1905.04499","2308.06880","2407.06424","2503.23253","2505.06813","2507.12829"]
for aid in ids:
    url = f"http://export.arxiv.org/api/query?id_list={aid}"
    try:
        r = requests.get(url, timeout=20)
        print(aid, r.status_code, r.text[:120].replace("\n"," "))
    except Exception as e:
        print("ERR", aid, e)
```

```python
import requests, textwrap, re, xml.etree.ElementTree as ET
ids = ["1501.02835","1905.04499","2308.06880","2407.06424","2503.23253","2505.06813","2507.12829"]
for aid in ids:
    url = f"http://export.arxiv.org/api/query?id_list={aid}"
    try:
        r = requests.get(url, timeout=20)
        print(aid, r.status_code, r.text[:160].replace("\n"," "))
    except Exception as e:
        print("ERR", aid, e)
```

```python
import requests, textwrap, re, xml.etree.ElementTree as ET
ids = ["1501.02835","1905.04499","2308.06880","2407.06424","2503.23253","2505.06813","2507.12829"]
for aid in ids:
    url = f"http://export.arxiv.org/api/query?id_list={aid}"
    try:
        r = requests.get(url, timeout=20)
        print(aid, r.status_code, r.text[:160].replace("\n"," "))
    except Exception as e:
        print("ERR", aid, e)
```

```python
import requests, textwrap, re, xml.etree.ElementTree as ET
ids = ["1501.02835","1905.04499","2308.06880","2407.06424","2503.23253","2505.06813","2507.12829"]
results=[]
for aid in ids:
    url = f"http://export.arxiv.org/api/query?id_list={aid}"
    try:
        r = requests.get(url, timeout=20)
        results.append((aid,r.status_code,r.text[:200].replace("\n"," ")))
    except Exception as e:
        results.append((aid,"ERR",str(e)))
results
```

Cactus flower moduli spaces are a family of closely related genus-\(0\) moduli spaces in which stable marked curves acquire a distinctly “cactus-like” or “flower-like” geometry. In one foundational incarnation, the real loci \(\overline{\mathcal M}_{0,n}(\mathbb R)\) of Deligne–Mumford compactifications parametrize stable rational curves over \(\mathbb R\), hence trees of circles with labeled points; their fundamental groups are pure cactus groups, and their topology, cohomology, and representation theory have been studied by \(FI\)-module methods, operadic models, and geometric group theory [1501.02835]. In a second, additive incarnation, the compactifications \(\overline F_n\) of \(F_n=(\mathbb C^n\setminus\Delta)/\mathbb C\) parametrize “flower curves,” namely bouquets of framed genus-\(0\) components glued at a common distinguished point, and arise as special fibers of one-parameter degenerations whose generic fibers are \(\overline M_{n+2}\) [2308.06880]. Weighted Hassett analogues over \(\mathbb R\) interpolate between the usual real Deligne–Mumford spaces and \(\mathbb{RP}^{n-2}\), producing weighted cactus groups and polytopal decompositions by products of permutahedra [2503.23253]. Taken together, these spaces form a nexus between real algebraic geometry, \(K(\pi,1)\)-topology, operad theory, and the representation theory of Gaudin models [2407.06424].

## 1. Geometric definitions and the “cactus” and “flower” pictures

For \(n\ge 3\), \(\overline{\mathcal M}_{0,n}\) is the smooth projective variety parametrizing stable genus-\(0\) curves with \(n\) labeled marked points. A stable rational curve \(C\) with marked points \(z_1,\dots,z_n\) is a finite union of projective lines \(C_1,\dots,C_k\) such that each marked point lies on exactly one component, any two components intersect transversely in at most one point, the dual graph is a tree, and each component carries at least three special points, marked or nodal. Over \(\mathbb R\), each \(\mathbb P^1\) is a topological circle, so the real locus \(M_n:=\overline{\mathcal M}_{0,n}(\mathbb R)\) may be visualized as a tree of circles decorated by labeled points; this is the geometric source of the cactus imagery [1501.02835].

A distinct but related construction begins from
\[
F_n=(\mathbb C^n\setminus\Delta)/\mathbb C,
\]
the configuration space of \(n\) distinct ordered points on the affine line modulo simultaneous translation. Its compactification \(\overline F_n\) parametrizes cactus flower curves: bouquets \(C=C_1\cup\cdots\cup C_m\) of framed cactus curves glued together at their distinguished points, collectively carrying \(n\) marked points distributed among the petals. Each petal is itself a framed stable nodal \(\mathbb P^1\) with a nonzero tangent vector at the distinguished point. This framing removes automorphisms of each component, so \(\overline F_n\) is a projective variety with a natural \(S_n\)-action permuting the marked points [2407.06424].

The additive compactification is controlled by coordinates \(\delta_{ij}=x_i-x_j\) on \(\mathfrak t_n=\mathbb C^n/\mathbb C\), extended projectively so that \(\delta_{ij}=\infty\) records points lying on different petals. The compactification \(\overline{\mathfrak t}_n\) is defined by
\[
\delta_{ij}+\delta_{jk}=\delta_{ik},\qquad \delta_{ij}+\delta_{ji}=0,
\]
equivalently by reciprocal coordinates \(\nu_{ij}\) satisfying
\[
\nu_{ij}\nu_{jk}=\nu_{ik}\nu_{jk}+\nu_{ij}\nu_{ik},\qquad \nu_{ij}+\nu_{ji}=0.
\]
The map \(\gamma:\overline F_n\to\overline{\mathfrak t}_n\) has fibers that are products of Deligne–Mumford spaces, making precise the idea that a flower curve is an additive configuration together with stable internal structure on each petal [2308.06880].

This terminology is not uniform across the literature. In some papers, “cactus flower” refers to the real loci of stable genus-\(0\) curves, especially when the tree-of-circles picture is emphasized; in others, it refers specifically to the additive compactifications \(\overline F_n\). The common feature is the replacement of a single irreducible \(\mathbb P^1\) by a tree or bouquet of components organized around a distinguished attachment pattern.

## 2. Real Deligne–Mumford loci, asphericity, and pure cactus groups

The real spaces \(\overline{\mathcal M}_{0,n}(\mathbb R)\) are compact connected smooth manifolds. In the indexing \(M_n=\overline{\mathcal M}_{0,n}(\mathbb R)\), one has \(\dim M_n=n-3\), with \(M_3\) a point, \(M_4\cong S^1\), and \(M_5\) a connected sum of five real projective planes [1501.02835]. In the shifted indexing \(\overline M_{0,n+1}(\mathbb R)\), the dimension is \(n-2\), so \(\overline M_{0,4}(\mathbb R)\cong \mathbb{RP}^1\cong S^1\) and \(\overline M_{0,5}(\mathbb R)\) is a compact connected non-orientable surface with Poincaré polynomial \(1+4t\) [1905.04499].

A basic theorem due to Davis–Januszkiewicz–Scott identifies these spaces as Eilenberg–MacLane spaces \(K(\pi,1)\). Their fundamental groups are pure cactus groups:
\[
PC_n:=\pi_1(\overline{\mathcal M}_{0,n}(\mathbb R)),
\]
or, in alternate notation,
\[
PJ_n:=\pi_1(\overline M_{0,n+1}(\mathbb R)),\qquad PCact_n=\pi_1(\overline M_{0,n+1}(\mathbb R)).
\]
Thus
\[
H^*(PC_n;\mathbb Q)\cong H^*(M_n;\mathbb Q).
\]
There is also an \(S_n\)-equivariant fundamental group, the cactus group, fitting into
\[
1\longrightarrow PJ_n\longrightarrow J_n\longrightarrow S_n\longrightarrow 1.
\]
Its standard generators \(s_{p,q}\), \(1\le p<q\le n\), satisfy involutivity, commutation on disjoint intervals, and the cactus conjugation relation on nested intervals [2503.23253].

The relation with configurations on the circle is explicit:
\[
M_{0,n+1}(\mathbb R)\cong \mathrm{Conf}_{n+1}(S^1)/\mathrm{PSL}_2(\mathbb R).
\]
A loop reversing the block of labels in an interval \([p,q]\) yields the generator \(s_{p,q}\), and van Kampen relations among such interval-reversal loops recover the cactus relations. This produces a geometric interpretation of \(PJ_n\) as the fundamental group of compactified configuration spaces of labeled points on \(S^1\), and for \(n=3,4\) one obtains explicit low-dimensional models: \(PJ_3\cong \pi_1(\overline X(4))\) with \(\overline X(4)\cong S^1\), and \(PJ_4\cong \pi_1(\overline X(5))\) with \(\overline X(5)\cong \#_5\,\mathbb{RP}^2\) [2505.06813].

These spaces also carry a canonical cellular decomposition by associahedra. Devadoss’ mosaic operad identifies \(\overline M_{0,n+1}(\mathbb R)\) as a quotient of the associahedral cell structure by local \(\mathbb Z_2\)-reflections; equivalently, cells are indexed by planar rooted trees modulo reflections at vertices. This operadic structure is compatible with gluing of stable curves and underlies many later constructions [1905.04499].

## 3. Cohomology rings, operadic models, and representation stability

The cohomology of the real Deligne–Mumford loci is exceptionally explicit. Etingof–Henriques–Kamnitzer–Rains identify \(H^*(M_n;\mathbb Q)\) with a skew-commutative algebra \(\Lambda_n\) generated in degree \(1\) by antisymmetric symbols \(w_{ijkl}\), \(1\le i,j,k,l\le n\), subject to
\[
w_{ijkl}+w_{jklm}+w_{klmi}+w_{lmij}+w_{mijk}=0,
\]
and
\[
w_{ijkl}w_{ijkm}=0.
\]
The \(S_n\)-action is induced by permuting indices,
\[
\pi\cdot w_{ijkl}=w_{\pi(i)\pi(j)\pi(k)\pi(l)}.
\]
A convenient generating set for \(H^1(M_n;\mathbb Q)\) is \(\{w_{1ijk}\mid 1<i<j<k\le n\}\), with no additional linear relations, and
\[
H^1(M_n;\mathbb Q)\cong \bigwedge^3\mathcal H_n,
\]
where \(\mathcal H_n\) is the standard \((n-1)\)-dimensional representation of \(S_n\). For \(n\ge 4\),
\[
H^1(M_n;\mathbb Q)\cong V(1,1,1)_n,
\]
and its character is the degree-\(3\) character polynomial
\[
\chi_{H^1(M_n;\mathbb Q)}=\binom{X_1}{3}+X_3-X_2X_1-\binom{X_1}{2}+X_2+X_1-1.
\]
In particular,
\[
\dim H^1(M_n;\mathbb Q)=\frac{(n-1)(n-2)(n-3)}{6}.
\]
Because the ring is generated multiplicatively by \(H^1\), these degree-\(1\) generators control all higher cohomology [1501.02835].

The \(FI\)-module formalism packages the sequence \(\{H^i(M_n;\mathbb Q)\}_n\) into functorial \(S_n\)-representations via forgetful morphisms
\[
\phi_f:\overline{\mathcal M}_{0,n}\to \overline{\mathcal M}_{0,m}
\]
along injections \(f:[m]\hookrightarrow [n]\). In this setting, \(H^i(M_\bullet;\mathbb Q)\) is generated in degree \(\le 3i+1\), has weight \(\le 3i\), and has stability degree \(\le 3i\). Therefore the Church–Ellenberg–Farb theorem yields uniform representation stability with stable range
\[
n\ge 6i.
\]
For \(n\ge 6i\), the character of \(H^i(M_n;\mathbb Q)\) is given by a unique character polynomial \(P_i\) in \(X_1,\dots,X_{3i}\) of degree \(\le 3i\), and the Betti number \(b_i(M_n)=\dim H^i(M_n;\mathbb Q)\) is given by a polynomial in \(n\) of degree \(\le 3i\) [1501.02835].

Operadically, the real loci form the mosaic operad. Khoroshkin–Willwacher describe the cellular chains as a cobar construction:
\[
\mathrm{Chains}_{\mathrm{cell}}(\overline M_{0,\bullet}(\mathbb R))\simeq \Omega([As]^\vee_{\mathbb Z_2}),
\]
and identify a homotopy quotient presentation by the \(\mathbb Z_2\)-action on the associative operad. The homology operad is isomorphic to the odd Poisson operad, while the cohomology rings are quadratic Koszul algebras generated by degree-\(1\) classes \(V_{ijk}\) with Arnol'd-type relations. The same work shows that \(\overline M_{0,n+1}(\mathbb R)\) is not formal for \(n\ge 6\), although it remains a rational \(K(\pi,1)\)-space. Thus asphericity does not imply formality in this setting [1905.04499].

A frequent misconception is that the real moduli spaces behave exactly like configuration spaces of points in the plane. The comparison with pure braid groups is strong but incomplete: the forgetful maps \(M_n\to M_{n-1}\) are not fibrations, and for \(n>5\) the spaces are not formal. The cactus analogy is therefore geometric and representation-theoretic, not a literal transfer of all braid-group phenomena [1501.02835].

## 4. Weighted Hassett spaces and weighted cactus groups

A weighted extension is obtained by replacing ordinary stability with Hassett \(\mathcal A\)-stability. For
\[
\mathcal A=(a_1,\dots,a_{n+1})\in(0,1]^{n+1},\qquad \sum_i a_i>2,
\]
an \((n+1)\)-pointed prestable curve \((C;x_\bullet)\) is \(\mathcal A\)-stable if for every smooth point \(p\in C\),
\[
\sum_{x_i=p} a_i\le 1,
\]
and for every irreducible component \(C'\subset C\),
\[
\#\{\text{nodes on }C'\}+\sum_{x_i\in C'}a_i>2.
\]
The paper on weighted real stable curves specializes to the \(S_n\)-symmetric vector
\[
\mathcal A(a)=\left(\frac1a,\dots,\frac1a,1\right),\qquad a\in\{1,2,\dots,n-1\}.
\]
Up to \(a\) of the first \(n\) points may collide, while the last point never collides. When \(a=1\) or \(a=2\), one recovers \(\overline M_{0,n+1}(\mathbb R)\); when \(a=n-1\) and \(n\ge 3\), one gets \(\mathbb{RP}^{n-2}\) [2503.23253].

The corresponding equivariant and ordinary fundamental groups are
\[
J_n^a:=\pi_1^{S_n}(\overline M_{0,\mathcal A(a)}(\mathbb R)),\qquad PJ_n^a:=\pi_1(\overline M_{0,\mathcal A(a)}(\mathbb R)).
\]
For \(a\ge 3\), \(J_n^a\) has generators \(s_{p,q}\) for intervals of length \(2\) or at least \(a+1\), satisfying the cactus relations together with braid relations
\[
(s_{i,i+1}s_{i+1,i+2})^3=1,\qquad 1\le i\le n-2.
\]
There is a short exact sequence
\[
1\longrightarrow PJ_n^a\longrightarrow J_n^a\longrightarrow S_n\longrightarrow 1,
\]
and the blowdown \(\overline M_{0,n+1}\to \overline M_{0,\mathcal A(a)}\) induces a surjection
\[
J_n\twoheadrightarrow J_n^a.
\]
As \(a\) increases, more cactus generators collapse to permutations, producing a chain of quotients
\[
J_n=J_n^1=J_n^2\twoheadrightarrow J_n^3\twoheadrightarrow\cdots\twoheadrightarrow J_n^{n-1}=S_n\rtimes \mathbb Z/2\mathbb Z.
\]
At the extreme \(a=n-1\), the \(\mathbb Z/2\)-factor is generated by the full reversal \(w_{1,n}\) [2503.23253].

The geometry of these weighted spaces is encoded by a dual polytopal decomposition. In the unweighted case, dual cells are cubes
\[
\overline{W_\tau}\cong [-1,1]^{|E_{\mathrm{int}}(\tau)|}
\]
indexed by stable trees \(\tau\). In the weighted case, for an \(a\)-stable tree with \(d\) internal edges and leaf labels \(A_1,\dots,A_r\), the corresponding dual cell is
\[
\overline{W_\tau}\cong [-1,1]^d\times \prod_{i=1}^r \Pi_{|A_i|},
\]
where \(\Pi_k\) is the permutahedron. Hexagonal \(\Pi_3\)-cells supply the braid relations; square cells supply commuting and cactus relations. This decomposition generalizes the known dual cube decomposition of \(\overline M_{0,n+1}(\mathbb R)\) and makes visible how weighted collisions deform the topology from cubes to products of permutahedra [2503.23253].

## 5. Additive flower curves, compactifications, and degeneration from \(\overline M_{n+2}\)

The additive spaces \(\overline F_n\) are natural compactifications of configurations on the affine line modulo translation. They are defined so that \(F_n\) is an open dense subset of dimension \(n-1\), and there is a natural map
\[
\overline F_n\to \overline{\mathfrak t}_n
\]
whose fiber over a point \(\delta\in\overline{\mathfrak t}_n\) is a product \(\prod_j \overline M_{|B_j|+1}\), where \(B=(B_1,\dots,B_m)\) records which marked points lie on each petal. Boundary strata are combinatorial: codimension-\(1\) strata arise either by separating marked points into two blocks \(I\cup J\) with \(\nu_{ji}=0\) for all \(i\in I\), \(j\in J\), or by collapsing a subset \(I\) through the conditions \(\delta_{ij}=0\) for \(i,j\in I\). Higher codimension strata are obtained by iteration [2407.06424].

A central structural result is the existence of a one-parameter degeneration from multiplicative to additive moduli. There is a group scheme over \(\mathbb A^1\),
\[
\mathbb A=\{(z;\varepsilon)\in\mathbb C^2\mid 1-\varepsilon z\neq 0\},
\]
with multiplication
\[
(z_1;\varepsilon)\cdot (z_2;\varepsilon)=(z_1+z_2-\varepsilon z_1z_2;\varepsilon).
\]
For \(\varepsilon\neq 0\), \(\mathbb A(\varepsilon)\cong \mathbb C^\times\), while \(\mathbb A(0)\cong\mathbb C\). The relative compactification \(\overline{\mathcal F}_n\to \mathbb A^1\) has fiber \(\overline M_{n+2}\) for \(\varepsilon\neq 0\) and fiber \(\overline F_n\) for \(\varepsilon=0\). Geometrically, two special marked points of a stable genus-\(0\) curve coalesce to form a single distinguished point with a nonzero tangent vector, yielding a flower curve in the limit [2407.06424].

The space \(\overline F_n\) admits affine charts \(W_\tau\) indexed by planar binary forests \(\tau\), equipped with regular functions \(\nu_{ij}\), \(\delta_{ij}\), and mixed terms \(\delta_{ij}\nu_{kl}\). It also contains distinguished loci: the equations \(\delta_{ij}=\infty\) for all \(i\ne j\) define a unique \(S_n\)-fixed maximal flower point \(\infty\), while \(\delta_{ij}=0\) for all \(i\ne j\) defines a divisor naturally isomorphic to \(\overline M_{n+1}\). More generally, in the total family over \(\mathbb A^1\), the locus \(\{\delta_{ij}=0\}\) is a divisor isomorphic to \(\overline M_{n+1}\times \mathbb A^1\) [2507.12829].

Over \(\mathbb R\), the space \(\overline F_n(\mathbb R)\) has an explicit combinatorial model by a \(\mathrm{CAT}(0)\) cube complex \(\widehat D_n\) indexed by unordered planar forests, and \(\overline{\mathfrak t}_n(\mathbb R)\) is modeled by a quotient \(\widehat P_n\) of the permutahedron obtained by identifying parallel faces. Both are nonpositively curved and hence aspherical. Their equivariant fundamental groups are
\[
\pi_1^{S_n}(\overline{\mathfrak t}_n(\mathbb R))\cong vS_n,\qquad
\pi_1^{S_n}(\overline F_n(\mathbb R))\cong vC_n,
\]
the virtual symmetric and virtual cactus groups. Moreover, a twisted real degeneration from \(\overline M_{n+2}^\sigma(\mathbb R)\) to \(\overline F_n(\mathbb R)\) induces a natural surjective homomorphism
\[
\widetilde{AC}_n\longrightarrow vC_n
\]
from the extended affine cactus group [2308.06880].

These additive spaces should not be conflated with the real loci \(\overline M_{0,n+1}(\mathbb R)\). The former compactify affine configurations modulo translation and carry virtual cactus symmetry; the latter are real Deligne–Mumford spaces with ordinary cactus symmetry. The degeneration \(\overline M_{n+2}\rightsquigarrow \overline F_n\) explains their proximity while preserving the distinction.

## 6. Operadic coverings, Gaudin models, and monodromy

The spaces \(\overline F_n\) and \(\overline M_{n+1}\) support compatible gluing maps. Besides the Deligne–Mumford operad structure \(\gamma_k\) on \(\{\overline M_{n+1}\}\), there is a right module structure
\[
\beta_k:\overline F_{k+1}\times \overline M_{n-k+1}\to \overline F_n
\]
and an internal flower-gluing operation
\[
\alpha_k:\overline F_k\times \overline F_{n-k}\to \overline F_n.
\]
Boundary divisors of \(\overline F_n\) are unions of images of the \(\alpha_k\) and \(\beta_k\), and \(\beta_0\) identifies \(\overline M_{n+1}\) as the divisor \(\delta\equiv 0\) in \(\overline F_n\). This yields a precise operadic framework in which finite coverings of real loci can be organized by tensor-categorical data [2507.12829].

Kamnitzer–Rybnikov show that isomorphism classes of operadic coverings of \(\overline F_n(\mathbb R)\) are naturally equivalent to equivalence classes of concrete coboundary monoidal categories satisfying semisimplicity and finiteness conditions. A \(\Xi\)-colored operadic covering consists of \(S_n\)-equivariant coverings
\[
E_n\to \overline F_n(\mathbb R)\times \Xi^n,\qquad
X_{n+1}\to \overline M_{n+1}(\mathbb R)\times \Xi^{n+1},
\]
together with covering isomorphisms compatible with \(\alpha_k\), \(\beta_k\), and \(\gamma_k\). On the categorical side, one has a concrete \(\Xi\)-colored coboundary monoidal category with multiplicity sets, associator, coboundary commutor, and faithful monoidal functor to sets. The equivalence reconstructs one structure from the other by parallel transport along canonical real paths in \(\overline F_2(\mathbb R)\) and \(\overline F_3(\mathbb R)\) [2507.12829].

The principal representation-theoretic application is to Gaudin algebras. For a simple Lie algebra \(\mathfrak g\), the homogeneous Gaudin algebra \(A(z)\subset (U\mathfrak g^{\otimes n})^{\mathfrak g}\) extends over \(\overline M_{n+1}\), the trigonometric Gaudin algebra \(A_\theta^{\mathrm{trig}}(z)\subset (U\mathfrak g^{\otimes n})^{\mathfrak h}\) extends over \(\overline M_{n+2}\), and the inhomogeneous Gaudin algebra \(A_\chi(z)\subset (U\mathfrak g^{\otimes n})^{\mathfrak z_{\mathfrak g}(\chi)}\) extends over \(\overline F_n\). These families are unified by
\[
A_\chi(z;\varepsilon)=
\begin{cases}
A^{\mathrm{trig}}_{\varepsilon^{-1}\chi}(1-\varepsilon z_1,\dots,1-\varepsilon z_n), & \varepsilon\ne 0,\\
A_\chi(z_1,\dots,z_n), & \varepsilon=0,
\end{cases}
\]
with
\[
\lim_{\varepsilon\to 0}A^{\mathrm{trig}}_{\varepsilon^{-1}\chi}(1-\varepsilon z_1,\dots,1-\varepsilon z_n)=A_\chi(z_1,\dots,z_n).
\]
Thus \(\overline{\mathcal F}_n\) is the universal parameter space for the degeneration from trigonometric to inhomogeneous Gaudin subalgebras [2407.06424].

Under explicit split or compact real conditions, these commutative algebras act with simple spectrum on tensor products of irreducible \(\mathfrak g\)-modules. The resulting sets of Bethe eigenlines form equivariant coverings over the relevant real loci. For inhomogeneous Gaudin algebras over \(\overline F_n(\mathbb R)\), the monodromy group is \(vC_n\); for trigonometric Gaudin algebras over the split real form of \(\overline M_{n+2}\), the monodromy group is the mirabolic cactus group; for the compact real form, it is the extended affine cactus group. In both trigonometric cases, the monodromy factors through \(vC_n\). Using the operadic-covering classification, one recovers Kashiwara \(\mathfrak g\)-crystals as a concrete coboundary category from these Bethe eigenline coverings, and the resulting \(vC_n\)-action matches the crystal commutor action on tensor products [2507.12829].

In the minuscule case, this provides a combinatorial version of the Bezrukavnikov–Okounkov wall-crossing conjecture for quantum cohomology of minuscule resolutions of affine Grassmannian slices. The point established in the source is precise: the eigenline monodromy of trigonometric Gaudin models over the compact real locus realizes the expected cactus-type wall-crossing on crystals [2507.12829].

## 7. Structural scope, comparisons, and open questions

Several adjacent moduli theories meet here. The real Deligne–Mumford spaces \(\overline M_{0,n+1}(\mathbb R)\) form the classical cactus landscape: they are smooth aspherical manifolds, mosaic-operadic, and governed by pure cactus groups. The weighted Hassett spaces \(\overline M_{0,\mathcal A(a)}(\mathbb R)\) interpolate between these and projective space, replacing cubes by products of permutahedra and introducing braid relations into cactus presentations. The additive compactifications \(\overline F_n\) are not merely weighted variants; they arise from a different quotient, \((\mathbb C^n\setminus\Delta)/\mathbb C\), and are tied to virtual cactus symmetry and to the degeneration of \(\overline M_{n+2}\) [2503.23253].

The comparison with braid-type families is pervasive but quantitatively different. Configuration spaces \(\mathcal F(\mathbb C,n)\), pure braid groups, pure virtual and flat braid groups, pure string motions, and \(\mathcal M_{0,n+1}\) all admit \(FI\)-module approaches when cohomology is generated in degree \(1\). The cactus case mirrors this pattern, but its generators have arity \(4\), namely the classes \(w_{ijkl}\), which is why weight and character-degree bounds are \(3i\) rather than the order-\(2i\) bounds typical in braid-type examples [1501.02835].

The main unresolved issues recorded in the cited work concern coefficients, group theory, and sharpness. The \(FI\)-module arguments require characteristic \(0\), usually \(\mathbb Q\), and do not address torsion in \(H^*(PC_n;\mathbb Z)\). It is not known whether \(PC_n\) is residually nilpotent. The effective stable ranges \(n\ge 6i\) and degree bounds \(\le 3i\) are not claimed to be optimal. On the additive side, \(\overline F_n\) has rich real topology and strong applications to Gaudin monodromy, but its deeper birational and singularity-theoretic features are not the focus of the cited sources [1501.02835].

A further source of ambiguity is that “cactus flower” has also been used in a different algebro-geometric context for low-socle loci inside Hilbert schemes that control Grassmann cactus varieties. There the relevant moduli functor \(M_{\ell,\le s}(\mathbb P^n)\) parametrizes degree-\(\ell\) finite subschemes with socle dimension at most \(s\), and the term denotes parameter spaces of low-socle finite schemes rather than genus-\(0\) nodal curves [2507.21586]. This suggests that the phrase has become a broader organizing label for moduli spaces governed by cactus-type incidence phenomena, but the curve-theoretic spaces \(\overline M_{0,n+1}(\mathbb R)\), \(\overline M_{0,\mathcal A(a)}(\mathbb R)\), and \(\overline F_n\) remain the central objects in the topology–operad–Gaudin literature.

Within that literature, cactus flower moduli spaces provide a unified geometric language for trees of circles, bouquets of framed rational components, weighted collision patterns, and degenerating Gaudin parameter spaces. Their significance lies in the simultaneous presence of explicit combinatorics, computable cohomology, operadic factorization, and nontrivial monodromy actions by cactus, affine cactus, and virtual cactus groups.

Source: https://www.emergentmind.com/topics/cactus-flower-moduli-spaces