---
title: Linked Projective Spaces
url: https://www.emergentmind.com/topics/linked-projective-spaces
type: topic
---

# Linked Projective Spaces

Linked projective spaces designate, in the most direct sense, projective configurations in which one projective structure is represented, organized, or reconstructed inside another. In Blunck and Havlicek’s treatment, the projective line over a ring \(R\) is represented inside an ordinary projective space \(\mathbb P(K,U\times U)\) by subspaces \(U^{(\rho_a,\rho_b)}\), thereby linking ring geometry to classical projective geometry [1304.0098]. In a later and formally distinct usage, linked projective spaces are quiver Grassmannians of constant dimension one attached to linked nets over \(\mathbb Z^n\)-quivers [2508.14799]. Closely related work studies complements of a fixed subspace inside a projective space, equips them with an affine structure, and interprets its lines as affine reguli or cones over affine reguli [1304.0093]. Taken together, these developments place the topic at the intersection of projective geometry over rings, Grassmannian-type parametrizations, affine structures on families of complements, and quiver-theoretic degeneration theory.

## 1. Terminological scope and conceptual profile

The phrase does not have a single universal meaning across the cited literature. In the ring-geometric setting, it refers to the passage from the projective line over a ring to a family of subspaces in an ambient projective space [1304.0098]. In the quiver-theoretic setting, linked projective spaces are a formally defined class of quiver Grassmannians associated with linked nets over \(\mathbb Z^n\)-quivers [2508.14799]. By contrast, “Affine Spaces within Projective Spaces” does not use the phrase literally, but studies a mathematically adjacent construction: the set of complements of a fixed subspace \(W\) in \(\mathbb P(K,V)\), organized as an affine space whose lines are affine reguli or cones over affine reguli [1304.0093].

This suggests that the expression functions less as a single canonical definition than as a family resemblance centered on three recurrent features: a distinguished ambient projective space, a parametrized family of subspaces inside it, and an incidence or complementarity relation that transfers algebraic data into projective geometry. In one direction, ring elements become endomorphisms and then graphs of operator pairs. In another, complements of a fixed subspace become points of an affine space. In a third, dimension-one subrepresentations of a quiver representation become embedded subschemes of a product of projective spaces.

A common misconception is that “linked projective spaces” refers simply to products of projective spaces. The cited works are more specific. The product construction matters in some contexts, but the salient structure is the compatibility law imposed on subspaces, complements, or lines. In [1304.0098], linkage is controlled by a \((K,R)\)-bimodule and the distant relation. In [1304.0093], it is complementarity to a fixed subspace \(W\). In [2508.14799], it is quiver-theoretic compatibility encoded by a linked net and realized as a quiver Grassmannian.

## 2. Projective lines over rings and their representation in projective space

For a ring \(R\) with identity \(1\), explicitly allowing the trivial case \(1=0\), the projective line over \(R\) is defined by
\[
\mathbb P(R):=R(1,0)^{\mathrm{GL}_2(R)}.
\]
Equivalently,
\[
\mathbb P(R)=\{\,R(a,b)\subset R^2\mid (a,b)\ \text{admissible}\,\},
\]
where \((a,b)\in R^2\) is admissible if there exist \(c,d\in R\) such that
\[
\begin{pmatrix} a & b\\ c & d\end{pmatrix}\in \mathrm{GL}_2(R).
\]
The paper also recalls the weaker notion of unimodularity:
\[
(a,b)\in R^2 \text{ is unimodular if there are }x,y\in R\text{ such that }ax+by=1.
\]
Every admissible pair is unimodular; if \(R\) is commutative, the two notions coincide, but over noncommutative rings they may differ [1304.0098].

The representation-theoretic bridge begins with a \((K,R)\)-bimodule \(U\), where \(U\) is a left vector space over a field \(K\) and a unitary right \(R\)-module satisfying
\[
k(u\cdot a)=(ku)\cdot a \qquad(k\in K,\ u\in U,\ a\in R).
\]
This is equivalent to a \(K\)-linear representation
\[
\varphi:R\to \operatorname{End}_K(U), \qquad a\mapsto \rho_a,
\]
where
\[
\rho_a:U\to U,\qquad u\mapsto u\cdot a.
\]

If \(S=\operatorname{End}_K(U)\), there is a bijection
\[
\Psi:\mathbb{P}(S)\to \mathcal{G}, \qquad S(\alpha,\beta)\mapsto U^{(\alpha,\beta)}:=\{(u^\alpha,u^\beta)\mid u\in U\},
\]
where \(\mathcal G\) is the set of all subspaces of \(\mathbb P(K,U\times U)\) that are isomorphic to one of their complements. Composing the induced map of projective lines with \(\Psi\) yields the projective representation
\[
\Phi:=\bar\varphi\,\Psi:\mathbb{P}(R)\to \mathcal{G}, \qquad R(a,b)\mapsto U^{(\rho_a,\rho_b)}.
\]
Explicitly,
\[
U^{(\rho_a,\rho_b)}=\{(u^{\rho_a},u^{\rho_b})\mid u\in U\}=\{(u\cdot a,u\cdot b)\mid u\in U\}\subset U\times U.
\]

This is the core ring-to-projective-space construction. Points of \(\mathbb P(R)\) become projective subspaces in \(\mathbb P(K,U\times U)\), and the image lies in the distinguished family \(\mathcal G\) of subspaces isomorphic to one of their complements. In this way, the projective line over a possibly noncommutative ring is linked to ordinary projective geometry.

## 3. Distance, complementarity, and algebraic pathologies

The intrinsic geometry of \(\mathbb P(R)\) is governed by the distant relation
\[
\Delta:=\{(R(1,0),R(0,1))\}^{\mathrm{GL}_2(R)}.
\]
For
\[
p=R(a,b),\qquad q=R(c,d),
\]
with admissible representatives, one has
\[
p\triangle q \iff
\begin{pmatrix} a & b\\ c & d \end{pmatrix}\in \mathrm{GL}_2(R).
\]
By definition, this means exactly that \(p\) and \(q\) are complementary submodules of \(R^2\). Over a field, distinct points are distant; over rings, non-distant points can still have trivial intersection, may span all of \(R^2\) without being complementary, or may exhibit intermediate behavior [1304.0098].

The main geometric theorem of the representation is:
\[
p\triangle q \Rightarrow p^\Phi,\ q^\Phi\ \text{complementary},
\]
and
\[
\Phi\ \text{injective} \iff U\ \text{faithful} \iff \varphi\ \text{injective}.
\]
Thus distant points always map to complementary subspaces in \(\mathbb P(K,U\times U)\), while faithful bimodules yield genuine projective models.

The converse need not hold. Proposition 4.5 gives the exact criterion:
\[
\Phi\ \text{maps non-distant points to non-complementary subspaces}
\iff
\big(\rho_a\in \operatorname{Aut}_K(U)\Rightarrow a\in R^*\big)\ \forall a\in R.
\]
If some nonunit \(a\in R\) acts invertibly on \(U\), then \(R(1,0)\) and \(R(1,a)\) are non-distant in \(\mathbb P(R)\), but their images are complementary. The counterexample
\[
R=K[X],\qquad U=K(X)
\]
shows this phenomenon explicitly: multiplication by \(X\) is bijective on the field of fractions \(K(X)\), although \(X\notin R^*\).

Several ring-theoretic conditions control the pathology of \(\mathbb P(R)\). The ring \(R\) is Dedekind-finite if \(ab=1\Rightarrow ba=1\). Proposition 2.2 shows the equivalence of:
1. \(R\) is Dedekind-finite;
2. whenever \(R(a,b)\in \mathbb P(R)\), the pair \((a,b)\) is admissible;
3. no point of \(\mathbb P(R)\) is properly contained in another point of \(\mathbb P(R)\).

The paper also recalls that if \(R\) has stable rank \(2\), then every unimodular pair is admissible and \(R\) is Dedekind-finite; and if \(R\) is finite-dimensional over a subfield, then \(R\) has stable rank \(2\), hence is a GE\(_2\)-ring [1304.0098]. These conditions matter when studying quotients and induced models, because they influence admissibility, surjectivity of induced maps, and the faithfulness with which the distant relation is reflected in ambient geometry.

## 4. Sub-bimodules, quotients, and classical geometric configurations

A major strength of the representation is that module-theoretic operations become projective constructions. If \(U'\subset U\) is a sub-bimodule, with induced representation \(\Phi'\), then
\[
p^{\Phi'} = p^\Phi \cap (U'\times U')
\qquad\text{for each }p\in \mathbb{P}(R).
\]
If \(U=U'\oplus U''\), then
\[
p^\Phi = p^{\Phi'}\oplus p^{\Phi''}.
\]
For the quotient \(\widetilde U=U/U'\), with associated representation \(\widetilde\Phi\),
\[
p^{\widetilde\Phi}=p^\Phi + (U'\times U').
\]
Submodules therefore correspond to intersection with a fixed ambient subspace, direct-sum decompositions correspond to direct-sum decompositions of represented subspaces, and quotient modules correspond to join constructions.

The examples show that these formulas recover familiar configurations in finite-dimensional projective geometry and also expose genuinely ring-theoretic effects. In the regular representation \(U=R\), assuming that \(R\) contains \(K\) as a subfield,
\[
R(a,b)^\Phi = R(a,b)\subset R^2=U\times U.
\]
Here the projective line sits inside the projective space associated with \(R^2\), and distant points correspond exactly to complementary subspaces. If \(U\) is faithful and \(R\) is finite-dimensional over a subfield \(L\), then \(\Phi\) maps non-distant points to non-complementary subspaces. The field-of-fractions example \(R=K[X]\), \(U=K(X)\) is the key counterexample to this good behavior.

Representative examples are summarized below.

| Ring and bimodule | Image geometry | Structural feature |
|---|---|---|
| \(R=K,\ U=K^2=U_1\oplus U_2\) | regulus in \(3\)-space | lines joining corresponding points |
| \(R=K^n,\ U=R\) | \((n-1)\)-subspaces meeting each \(U_i\times U_i\) | coordinatewise separation |
| \(R=K[\varepsilon],\ \varepsilon^2=0\) | parabolic linear congruence when \(\alpha=\mathrm{id}\) | local-ring equivalence classes |
| upper triangular \(2\times 2\) matrix ring, \(U=K^2\) | non-pappian analogue of a special linear complex | lines meeting a fixed line |
| \(R=K[\varepsilon,\delta],\ \varepsilon^2=\delta^2=\varepsilon\delta=0\) | planes meeting a \(3\)-space in a regulus element | higher-dimensional local-ring model |

These examples display the range of the representation: reguli, hyperbolic linear congruences, special linear complexes, and local-ring foliations by equivalence classes all arise as projective models of \(\mathbb P(R)\). The paper also notes that \(\mathbb P(R)\) is the point set of a chain geometry \(\Sigma(F,R)\) or generalized chain geometry, but the main contribution is the representation of the point set itself rather than the full chain structure [1304.0098].

## 5. Affine structures on complements inside projective spaces

A complementary development starts from an ordinary projective space
\[
\mathbb P(K,V)
\]
with a fixed proper nonzero subspace
\[
W\le V.
\]
The central object is
\[
\mathcal S:=\{S\le V \mid V=W\oplus S\},
\]
the set of all complements of \(W\). Fixing one complement \(U\in\mathcal S\) with \(V=W\oplus U\), every complement can be written as
\[
S=U^{(\gamma,\delta)}:=\{u^\gamma+u^\delta\mid u\in U\},
\]
with \(\gamma\in \operatorname{Hom}(U,W)\) and \(\delta\in \operatorname{Aut}(U)\), hence also as
\[
S=U^{(\gamma,1)}.
\]
Thus \(\mathcal S\) is identified with \(\operatorname{Hom}(U,W)\) by
\[
U^{(\gamma,1)}\longleftrightarrow \gamma.
\]

To turn \(\mathcal S\) into an affine space, one chooses a basis \((b_i)_{i\in I}\) of \(U\) and defines an embedding
\[
\lambda:K\to \operatorname{End}(U),\qquad k\mapsto \lambda_k,
\]
where \(\lambda_k(b_i)=k\,b_i\). The induced scalar multiplication on \(\operatorname{Hom}(U,W)\) is
\[
k\cdot \gamma:=\lambda_k\gamma.
\]
Via the identification above,
\[
U^{(\gamma,1)}+U^{(\eta,1)} = U^{(\gamma+\eta,1)}, \qquad
k\cdot U^{(\gamma,1)} = U^{(\lambda_k\gamma,1)}.
\]
The resulting affine space is denoted
\[
\mathbb A(\mathcal S,(b_i)).
\]

This affine structure is generally not canonical. It depends on the choice of basis, equivalently on the chosen embedding \(K\hookrightarrow \operatorname{End}(U)\). If \(K\) is commutative, the affine structure becomes canonical. More generally, Theorem 2.6 states that two such affine structures coincide precisely when the associated projective \(Z\)-subspaces of \(\mathbb P(K,U)\), where \(Z\) is the center of \(K\), coincide.

Lines in this affine space have the form
\[
\ell(\alpha,\beta):=\{U^{(k\alpha+\beta,1)}\mid k\in K\}, \qquad \alpha,\beta\in\operatorname{Hom}(U,W),\ \alpha\neq 0.
\]
In the symmetric case \(W\cong U\), a line is regular if \(\alpha\in \operatorname{Aut}(U)\), and Proposition 3.1 states:
\[
\text{A line of }\mathbb A\text{ is regular } \Longleftrightarrow \text{ it is an affine regulus in }\mathbb P(K,V).
\]
In the general case, injective \(\alpha\) yields an affine regulus in a smaller projective subspace, while noninjective \(\alpha\) yields a cone over an affine regulus, with vertex the maximal central part of \(\ker(\alpha)\). If \(\ker(\alpha)\) is central, then
\[
\ell(\alpha,0)=\{X+\ker(\alpha)\mid X\in \mathcal R,\ X\neq \operatorname{im}(\alpha)\}.
\]

The framework is applied to dual spreads. A dual spread is a set of pairwise complementary subspaces such that each hyperplane contains one of its elements. Proposition 5.4 gives the criterion that \(\mathcal B\subseteq \mathcal S\), together with \(W\), is a dual spread iff:
\[
\text{(DS1)}\quad \text{distinct elements of }\mathcal{B}\text{ are joined by a regular affine line},
\]
\[
\text{(DS2)}\quad \text{each maximal singular subspace }\mathcal{S}(X)\text{ contains an element of }\mathcal{B}.
\]
Theorem 5.5 then characterizes all dual spreads containing \(W\) in terms of *-transversal families [1304.0093].

## 6. Quiver-theoretic linked projective spaces and the diagonal Chow class

In a formally defined later usage, linked projective spaces are quiver Grassmannians of constant dimension one attached to linked nets over \(\mathbb Z^n\)-quivers. A quiver \(Q\) with a partition \(T\) of its arrow set into \(n+1\) types is a \(\mathbb Z^n\)-quiver if: there is exactly one arrow of each type leaving each vertex; each vertex is connected to each other by an admissible path; and two paths leaving the same vertex arrive at the same vertex if and only if the difference of their type functions is constant. A representation \(V\) is a linked net if it is weakly linked and, for admissible paths \(\gamma_1,\gamma_2\) leaving the same vertex and having no arrow type in common,
\[
\ker(\varphi_{\gamma_1})\cap \ker(\varphi_{\gamma_2})=0.
\]
For a nontrivial finitely generated exact linked net, the minimum set of generators \(H\) is convex, meaning
\[
P(H)=H.
\]

If \(V\) is a representation in nonzero finite-dimensional vector spaces, the linked projective space \(LP(V)\) is the quiver Grassmannian of subrepresentations of pure dimension \(1\). For a minimum generating set \(H\), it is embedded as
\[
LP_H(V):=\left\lbrace (s_v \vert v\in H)\in \prod_{v\in H} P(V_v) \,\,\Bigg\vert\,\, \varphi^v_w(s_v)\wedge s_w=0 \text{ for all } v,w\in H\right\rbrace.
\]
Earlier work recalled in the paper states that \(LP(V)\) is generically smooth, a local complete intersection, and reduced. It was introduced as a tool for describing schematic limits of families of divisors.

The new contribution is combinatorial and intersection-theoretic. To each generator \(v\in H\), one associates a vector space \(W_v\subseteq \bigoplus_{u\in H}V_u\), modular data \((\mu_v,\mu_v^*)\), and a polymatroid base polytope \(P_{\mu_v}\). Theorem 4.2 states that the collection of polytopes \(P_{\mu_v}\) and their faces gives a polyhedral tiling of the standard simplex
\[
\Omega_{r+1},
\]
in particular
\[
P_V=\Omega_{r+1}.
\]
This polymatroidal tiling is the key combinatorial input.

Let
\[
P:=\prod_{v\in H}P(V_v),
\qquad
A^*(P)=\frac{\mathbb Z[h_v\, \vert\, v\in H]}{(h_v^{r+1}\vert v\in H)}.
\]
For a closed subscheme \(X\subseteq P\) of pure dimension \(r\),
\[
[X]=\sum_{q\in \Omega_r(\mathbb Z)} \deg_P^q(X)\prod_{v\in H} h_v^{r-q(v)}.
\]
The Chow class of the small diagonal in \(P\) is
\[
\sum_{q\in \Omega_r(\mathbb Z)} \prod_{v\in H} h_v^{r-q(v)}.
\]
Theorem 4.4 proves that linked projective spaces have exactly this class:
\[
[LP(V)]=\sum_{q\in \Omega_r(\mathbb Z)} \prod_{v\in H} h_v^{r-q(v)}.
\]
Thus linked projective spaces have the Chow class of the diagonal.

This does not settle the stronger degeneration problem. The paper states that it is an open question whether linked projective spaces are degenerations of the small diagonal. If the linked net \(V\) is smoothable, then \(LP(V)\) is the special fiber of a Mustafin variety and hence a flat degeneration of the small diagonal. In general, however, the diagonal Chow class is weaker than actual degeneration. This distinction is one of the main current points of interpretation and caution in the theory [2508.14799].

Source: https://www.emergentmind.com/topics/linked-projective-spaces