---
title: Semi-Stability in Loop Groups
url: https://www.emergentmind.com/topics/semi-stability-for-loop-groups
type: topic
---

# Semi-Stability in Loop Groups

Searching arXiv for the cited papers and related loop-group semistability work.
Semi-stability for loop groups is a numerical and geometric framework for the positive half of a real loop group, in which points of the loop symmetric space \(X=K\backslash G^+\) are tested against affine height functionals attached to rational parabolics and arithmetic cosets. In the form developed for untwisted affine Kac–Moody loop groups, it yields a notion of \(Q\)-semistability, a global semistable locus, a unique canonical destabilizing pair for each unstable point, and a stratification of both the symmetric space and its Borel–Serre type bordification [2507.13980]. In parallel, the loop-group literature also uses “stability” in a distinct Atiyah–Jones-type topological sense for moduli of holomorphic maps into loop-group flag manifolds and for associated holomorphic \(G\)-bundles with flag data along a line; that second usage concerns homology and homotopy stabilization rather than Harder–Narasimhan or GIT slope inequalities [0812.3684].

## 1. Affine loop-group setting

The basic objects are fixed by starting with a split simple Chevalley group over the reals, \(G_0\), with maximal compact subgroup \(K_0\), and then passing to the untwisted affine Kac–Moody algebra \(\widehat{\mathfrak g}\) built from its Lie algebra. On the group side one works with the “complete” Kac–Moody loop group \(G\), generated inside \(\operatorname{Aut}(V_\lambda)\) by exponentials of real root vectors and loop rotations \(\eta(s)\), together with the extended group \(G^e=\langle G,\eta(s)\mid s\in \mathbb R^\ast\rangle\). The torus \(A\subset T\subset G^e\) is attached to the extended Cartan \(\widehat{\mathfrak h}^e=\widehat{\mathfrak h}\oplus \mathbb RD\oplus \mathbb Rc\), where \(D\) is the degree derivation and \(c\) is the canonical central element [2507.13980].

The positive half of the real loop group is the semigroup
\[
G^+ := \bigcup_{r>0} G^r,\qquad G^r := G\cdot \eta(e^{-r}),
\]
with \(\eta(e^{-r})\) acting by contraction along the degree derivation \(D\). At the Lie algebra level, the relevant domain is the Tits cone
\[
\mathfrak h^+ = \{H-rD\mid H\in \mathfrak h,\ r>0\}
= \left\{\sum_{i\in I} p_i\,\check\omega_i + m\,c \;\middle|\; \sum_{i\in I} p_i d_i < 0\right\}.
\]
Its exponential is the positive torus slice \(A^+=\bigcup_{0<s<1} A^s\), and in coweight coordinates one has
\[
A^+ \cong \Big\{(s_1,\dots,s_{\ell+1};t)\in (\mathbb R_{>0})^{\ell+1}\times \mathbb R_{>0}\ \Big|\ s_1^{d_1}\cdots s_{\ell+1}^{d_{\ell+1}}<1\Big\}.
\]

The corresponding loop symmetric space is
\[
X:=K\backslash G^+.
\]
By Iwasawa decomposition there is a canonical homeomorphism
\[
X\simeq A^+\times U,
\]
where \(U\) is the pro-unipotent radical of the Borel subgroup \(B=T\ltimes U\). More generally, for each rational parabolic \(P\), the \(P\)-horospherical decomposition gives canonical coordinates
\[
X \simeq X_{M_P}\times A_P^+\times U_P,
\]
with \(M_P\) the Levi factor, \(A_P^+\subset A^+\) a subtorus slice, and \(U_P\) the pro-unipotent radical. These coordinates provide the height maps
\[
H_P:X\to (\mathfrak h^+)_{P},\qquad H_P^+:X\to \mathfrak h_P^+,
\]
which are the numerical input for semi-stability.

Rational parabolics are defined with respect to the arithmetic structure. If \(P_J=\bigcup_{w\in W(J)}BwB\) is a standard parabolic for a subset \(J\subset I\), then the rational parabolics are its \(G(\mathbb Z)\)-conjugates:
\[
\mathrm{Par} := \{gP_Jg^{-1}\mid g\in G(\mathbb Z),\ J\subset I\},\qquad \mathrm{Par}'=\mathrm{Par}\setminus\{G\}.
\]
In the affine setting one restricts to positive parabolic subsets.

## 2. Instability functions and the definition of semi-stability

Semi-stability is defined by minimizing affine height functionals over parabolics and arithmetic cosets. Let
\[
\rho:=\sum_{i\in I}\lambda_i\in \widehat{\mathfrak h}^{e\ast}
\]
be the affine analogue of the half-sum of positive roots. For a standard parabolic \(Q\subsetneq G\), set
\[
\rho_P^Q:=\rho(Q)|_{(\mathfrak h^+)_{Q}},\qquad \rho_P:=\rho-\rho(Q)\quad\text{on }(\mathfrak h^+)_{P},
\]
and define the degree of \(Q\)-instability of \(x\in X\) by
\[
\deg^{Q}_{\mathrm{inst}}(x):=\min_{P\subset Q}\;\min_{\delta\in Q(\mathbb Z)/P(\mathbb Z})}\;\big\langle\rho_P^Q,\;H_P^+(x\,\delta)\big\rangle.
\]
For \(Q=G\), one obtains
\[
\deg_{\mathrm{inst}}(x):=\min_{P\subset G}\;\min_{\delta\in G(\mathbb Z)/P(\mathbb Z})}\;\big\langle\rho_P,\;H_P^+(x\,\delta)\big\rangle.
\]
A point is \(Q\)-semistable if \(\deg^{Q}_{\mathrm{inst}}(x)\ge 0\), and semistable if \(\deg_{\mathrm{inst}}(x)\ge 0\) [2507.13980].

For \(Q\subsetneq G\), Proposition 5.2 gives an equivalent array of inequalities:
\[
\langle \omega, H_P^+(x\,\delta)\rangle \ge 0
\quad\text{for all }P\subset Q,\ \delta\in Q(\mathbb Z)/P(\mathbb Z}),\ \omega\in \widehat{\Delta}_P^Q.
\]
For maximal parabolics \(P_i\) and \(Q=G\), the criterion takes the form
\[
\langle\lambda_i, H_B(x\,\delta)\rangle \ge -\,r\,\frac{d_i}{\kappa_i}
\quad\text{for all }\delta\in G(\mathbb Z)/B(\mathbb Z}),
\]
where \(r\) is the \(D\)-slice, \(d_i\) comes from \(\rho(P_i)\), and \(\kappa_i>0\) is determined by
\[
\rho_{P_i}=\kappa_i\lambda_i-d_i\,\delta.
\]

A defining affine feature is that semi-stability is not invariant under the central subgroup \(A_c\). The paper records the shift formula
\[
\deg_{\mathrm{inst}}(\zeta x)=\langle \rho,H_c(\zeta)\rangle+\deg_{\mathrm{inst}}(x).
\]
This is one of the main departures from finite-dimensional Harder–Narasimhan theory: the central direction contributes an independent non-compact parameter, and semi-stability is sensitive to that parameter.

## 3. Canonical pairs and the semi-stability stratification

The affine analogue of the Harder–Narasimhan–Behrend–Chaudouard canonical reduction is formulated in terms of canonical pairs. For every \(x\in X\), the minimum defining \(\deg_{\mathrm{inst}}(x)\) exists. If \(\deg_{\mathrm{inst}}(x)<0\), there exists a unique destabilizing extremal pair \((P,\delta)\), with \(P\) a standard proper parabolic and \(\delta\in G(\mathbb Z)/P(\mathbb Z)\), such that
\[
\deg_{\mathrm{inst}}(x)=\langle \rho_P,H_P^+(x\,\delta)\rangle,
\]
and \(P\) is maximal by inclusion among all minimizers [2507.13980].

The characterization of a canonical pair has two parts. A pair \((P,\delta)\) is canonical for \(x\) if and only if:

1. The projection \(m:=\mathrm{pr}_{M_P}(x\,\delta)\in X_{M_P}\) is semistable for the finite-dimensional group \(M_P\).
2. The \(A\)-projection satisfies
   \[
   H_P^+(x\,\delta)\in A^+_{P,(0,1)}:=\{a\in A_P^+\mid \langle\alpha,\log a\rangle<0\ \forall \alpha\in \Delta_P,\ \langle\rho_P,\log a\rangle<0\}.
   \]

The extra inequality \(\langle\rho_P,\log a\rangle<0\) is specific to the affine setting. This condition is not reducible to the simple-root inequalities alone and is one of the structural novelties of loop-group semi-stability.

Canonical pairs produce a partition
\[
X = X^{\mathrm{ss}}\;\sqcup\;\bigsqcup_{P\in \mathrm{ParStd}'} (P),\qquad
(P):=\bigsqcup_{\delta\in G(\mathbb Z)/P(\mathbb Z})} (P,\delta),
\]
where \((P,\delta):=\{x\mid \mathrm{cp}(x)=(P,\delta)\}\). Each stratum carries a canonical fibration
\[
f_{(P,\delta)}:\ (P,\delta)\longrightarrow X_{M_P}^{\mathrm{ss}}\times A^+_{P,(0,1)},
\]
whose fibers are homeomorphic to \(U_P\). In the standard case,
\[
(P,e)=X_{M_P}^{\mathrm{ss}}\times A^+_{P,(0,1)}\times U_P.
\]

This structure mirrors the finite-dimensional pattern of numerical instability, canonical reduction, and Levi-factor induction, but it does so in a space with pro-unipotent radicals, a central affine direction, and rational parabolics indexed by an affine building rather than a spherical one.

## 4. Bordification, arithmetic quotients, and the affine building

The semi-stability theory is intertwined with a Borel–Serre type bordification of the loop symmetric space. For each rational parabolic \(P\), the attached boundary component is
\[
e(P):=X_{M_P}\times U_P,
\]
and the bordification is
\[
\overline X := X\;\cup\;\bigsqcup_{P\in \mathrm{Par}'} e(P).
\]
The corner attached to \(P\) is
\[
c(P):=X\;\cup\;\bigsqcup_{Q\supset P} e(Q).
\]
The embedding extension lemma provides a canonical open embedding
\[
\iota_P:\ X_{M_P}\times c(A_P^+)\times U_P\hookrightarrow c(P),
\]
and the topology is specified by Moore–Smith convergence classes. With this topology, \(\overline X\) is Hausdorff, and the natural right action of \(G(\mathbb Z)\) on \(X\) extends continuously to \(\overline X\) [2507.13980].

The boundary is indexed by rational parabolics and has the homotopy type of the affine, rational Tits building:
\[
\partial\overline X \simeq \mathcal T(G).
\]
Here \(\mathcal T(G)\) is the simplicial complex whose vertices are maximal rational parabolics and whose simplices are their finite intersections when those intersections are again parabolic. The homotopy equivalence is obtained from the nerve of the closed cover by the maximal boundary faces \(\overline{e(Q)}\).

The arithmetic group is defined by an integral structure on a highest-weight module:
\[
\Gamma := \{g\in G\mid g(V_{\lambda,\mathbb Z})\subset V_{\lambda,\mathbb Z})\}.
\]
Garland’s reduction theory supplies Siegel sets
\[
\mathfrak S_{t,\Omega}:=A_t^+\times U_\Omega,\qquad
A_t^+:=\{h\in A^+\mid h^{\check\alpha_i}<t,\ i\in I\},
\]
with the key property that every point of \(X\) can be translated into such a set by some \(\gamma\in \Gamma\). It also detects when intersections of truncated Siegel sets force \(\gamma\) into a proper parabolic, and it shows that sufficiently negative \(D\)-height places one near a parabolic wall.

Unlike the finite-dimensional Borel–Serre construction, the quotient \(\overline X/\Gamma\) is not compact. The boundary quotient \(\partial\overline X/\Gamma\) is compact, and the Levi and unipotent factors of parabolic ends have compact quotient, but the central \(A_c\)-direction remains non-compact. The non-compactness of the interior quotient is therefore traced precisely to the center of the loop group.

The semi-stability partition extends to the bordification. Completing the \(A\)-direction gives the bordified strata
\[
X\{P\}:=X_{M_P}^{\mathrm{ss}}\times c\bigl(A^+_{P,(0,1)}\bigr)\times U_P,
\qquad
X\{S\}:=R_{\delta^{-1}}\bigl(X\{P\}\bigr),
\]
and one obtains
\[
\overline X = X^{\mathrm{ss}}\;\sqcup\;\bigsqcup_{S\in \mathrm{Par}'} X\{S\}.
\]
This decomposition is compatible with the boundary faces: \(X\{S\}\cap e(Q)\neq\emptyset\) if and only if \(S\subset Q\).

## 5. Loop-group flags, holomorphic bundles, and the distinct topological notion of stability

A separate but closely related branch of the subject identifies loop-group flag manifolds with moduli of holomorphic bundles on rational curves and ruled surfaces. Let \(K\) be a compact semisimple Lie group with complexification \(G\), and let \(P_{LG}\subset LG\) be a standard parabolic. Then \(LG/P_{LG}\) parametrizes triples \((E,t,R)\), where \(E\) is a \(G\)-bundle on \(\mathbb P^1\), \(t\) is a trivialization on \(|z|\ge 1\), and \(R\) is a reduction to \(P_G\) over \(z=0\). In particular, \(LG/LG^+\) corresponds to pairs \((E,t)\), and for families over a curve \(\Sigma\), maps \(\operatorname{Hol}(\Sigma,LG/P_{LG})\) correspond to such triples on \(\Sigma\times \mathbb P^1\) [0812.3684].

For based maps from \(\mathbb P^1\), the relevant moduli are \(G\)-bundles on \(\mathbb P^1\times \mathbb P^1\) that are trivialized on
\[
\mathbb P^1\times \{z=\infty\}\ \cup\ \{x=\infty\}\times \mathbb P^1,
\]
with a \(P_G\)-reduction along \(z=0\). The Birkhoff and Bruhat stratifications of \(LG/P_{LG}\) then control degrees, divisors, and principal parts. The divisors \(Z_\lambda\) coming from lowest weight sections \(v_\lambda\) generate \(H_2(LG/P_{LG};\mathbb Z)\), and the multi-degree of a based map is measured by the intersection multiplicities with those divisors.

This geometric setup is related to parabolic bundle language, but not identical with the semi-stability theory above. The flag \(R\) along \(z=0\) is precisely a parabolic structure in the standard sense, and for \(G=GL_n\) the paper works with bundles equipped with flags at \(z=0\) via the nested sheaves \(E_{i,j}\). However, the paper does not develop the slope-based GIT semistability criterion for principal \(G\)-bundles; instead, its central “stability” results are topological stabilization theorems for spaces of based holomorphic maps.

The principal results are Atiyah–Jones-type statements. The stabilization map
\[
T_{k,k'}:\operatorname{Hol}_k(\mathbb P^1,LG/P_{LG})\to \operatorname{Hol}_{k'}(\mathbb P^1,LG/P_{LG})
\]
induces isomorphisms in homology \(H_j\) for
\[
j\le \big[\min(1/2,c(LG/P_{LG}))\,\ell(k)\big]-1,
\]
and the inclusion
\[
I:\operatorname{Hol}_k(\mathbb P^1,LG/P_{LG})\to \Omega_k^2(LG/P_{LG})
\]
induces isomorphisms in the same homology range and in homotopy groups \(\pi_j\) for
\[
j\le \big[\min(1/2,c(LG/P_{LG}))\,\ell(k)\big]-s-2.
\]
For instantons, Donaldson’s theorem identifies the moduli space \(M_{K,k}\) of framed \(K\)-instantons of charge \(k\) on \(S^4\) with \(\operatorname{Hol}_*^k(\mathbb P^1,LG/LG^+)\), giving a corresponding Atiyah–Jones theorem for \(M_{K,k}\). For \(SU(2)\) calorons, the moduli \(M_{SU(2),k_0,k_1}\) are identified with maps into \(LSL(2,\mathbb C)/B\), equivalently with rank-two bundles on \(\mathbb P^1\times \mathbb P^1\) with \(c_2=k_0\), specified framings, and a subline bundle over \(z=0\) of degree \(-k_1\).

The terminological distinction is essential. In the affine symmetric-space framework, semi-stability is a numerical condition defined by height minimization over rational parabolics. In the holomorphic-bundle framework of loop-group flags, “stability” refers to homology and homotopy stabilization of moduli spaces as the multi-degree grows.

## 6. Examples, finite-dimensional comparison, and open directions

In the affine \(\widehat{\mathfrak{sl}}_2\) case, one has \(I=\{1,2\}\) and
\[
A^+ \cong \{(s_1,s_2;t)\in (\mathbb R_{>0})^3\mid s_1s_2<1\}.
\]
For the maximal parabolics \(P_1\) and \(P_2\),
\[
\rho_{P_i}=\kappa_i\lambda_i-d_i\,\delta,\qquad i=1,2.
\]
Semi-stability for \(P_2\) reads
\[
\langle \lambda_2,H_B(x\,\delta)\rangle\ge 0,
\]
while for \(P_1\) one obtains the affine threshold
\[
\langle \lambda_1,H_B(x\,\delta)\rangle \ge r\,(d_1/\kappa_1).
\]
Canonical pairs are detected by translating \(x\) into the appropriate Siegel set and finding the unique maximal \(P\) for which \(\langle \rho_P,H_P^+(x\,\delta)\rangle<0\). In this rank-one case the fibers \(U_P\) are one-dimensional. The paper also exhibits a simple escaping sequence along the central direction, showing explicitly that \(X^{\mathrm{ss}}/\Gamma\) is non-compact [2507.13980].

For \(\widehat{\mathfrak{sl}}_n\), the maximal parabolics \(P_i\) correspond to the affine block structures, and canonical pairs \((P,\delta)\) are determined by the unique maximal block-structure \(P\) maximizing the negative height \(\langle \rho_P,H_P^+(x\,\delta)\rangle\). The associated stratum fibers over \(X_{M_P}^{\mathrm{ss}}\times A^+_{P,(0,1)}\) with fiber \(U_P\). This formulation makes the analogy with finite-dimensional canonical reduction especially transparent.

The comparison with finite-dimensional theory has two sides. The similarities are direct: there is a numerical criterion via height functionals, a minimization over parabolics and cosets, existence and uniqueness of a canonical destabilizing reduction, and a stratification by canonical pairs with fibrations to lower-rank semistable Levi spaces. The differences are equally structural: semi-stability is not invariant under the central direction \(A_c\), the affine condition \(\langle \rho_P,\cdot\rangle<0\) is genuinely new, and the bordification is only a partial compactification, with compact boundary quotient but non-compact total quotient because of the center.

Open directions identified in the recent affine theory include establishing Siegel finiteness in full generality for pro-unipotent radicals, comparing the semistability partition with Arthur’s truncations for affine groups, constructing \(\Gamma\)-equivariant retractions or tilings of \(\overline X\) analogous to finite-dimensional constructions, and extending the framework to adelic loop groups over global fields. A plausible implication is that these questions will determine how closely loop-group semi-stability can parallel both arithmetic reduction theory and the geometry of moduli of bundles on higher-dimensional bases.

Source: https://www.emergentmind.com/topics/semi-stability-for-loop-groups