Space of norms on locally algebraic representations

Technical Summary: Space of Norms on Locally Algebraic Representations

1. Problem Statement and Context
The paper investigates the geometry of the space of non-Archimedean norms on a locally algebraic representation $V = \pi_{\text{sm}} \otimes_E \sigma_{\text{alg}}$ of a reductive group $G = \mathbf{G}(F)$, where $F$ and $E$ are finite extensions of $\mathbb{Q}_p$. The representation $V$ is the tensor product of a smooth admissible representation $\pi_{\text{sm}}$ and a finite-dimensional algebraic representation $\sigma_{\text{alg}}$.

The central arithmetic problem motivating this work is the construction of a $G$-invariant norm on $V$. Such norms are crucial for integrality questions in the $p$-adic Langlands program, particularly regarding the Breuil–Schneider conjecture. While necessary conditions for the existence of such norms are known (e.g., Emerton's condition on Jacquet modules), the transition from these spectral conditions to the existence of a global invariant norm remains a significant obstacle. The paper aims to separate the general geometric framework governing these norms from the difficult arithmetic inputs required to satisfy the boundedness conditions.

2. Methodology and Geometric Framework
The author employs a metric-geometric approach, utilizing the extended Goldman–Iwahori distance $d_\infty$ on the space of norms.

• Projective Limit Structure: The space of norms on $V$ is identified with a projective limit of extended Bruhat–Tits buildings attached to finite-dimensional subspaces $V_K = \pi^K_{\text{sm}} \otimes_E \sigma_{\text{alg}}$. The author distinguishes between the ordinary projective limit (coordinatewise convergence) and the bounded projective limit (uniform convergence), arguing that the latter is the correct object for studying bounded orbits.
• Metric Geometry: The distance $d_\infty$ is shown to be an $\ell_\infty$-type metric, not a CAT(0) metric. The finite-distance component $N_{\alpha_0}(V)$ of a reference norm $\alpha_0$ is proven to be hyperconvex (injective). This property ensures that every bounded family of norms has a circumcenter with radius exactly half the diameter of the family.
• Fixed-Point Strategy: The paper avoids relying on CAT(0) fixed-point theorems. Instead, it uses the hyperconvexity of the space to construct a $G$-invariant norm directly as the pointwise supremum of a bounded orbit: $\alpha_G(v) = \sup_{g \in G} \alpha_0(g^{-1}v)$.

3. Key Contributions and Results

• Geometric Characterization of Invariant Norms (Theorem 1.1, Theorem 5.1):
  The paper establishes a precise equivalence: a finite-distance component $N_{\alpha_0}(V)$ contains a $G$-invariant norm if and only if the orbit $G\alpha_0$ is bounded with respect to the Goldman–Iwahori metric. If the orbit is bounded, the invariant norm is the pointwise supremum of the orbit, and the distance from the reference norm to the invariant locus is exactly half the orbit's diameter.

• Reduction to Finite-State Problems (Section 3 & 6):
  The existence of a bounded orbit is reduced to a "finite-state" problem. The author formulates a criterion (Theorem 3.2) where a $G$-invariant norm exists if and only if every finite building diagram (involving restriction and transport maps) has a solution. This allows the use of compactness arguments to lift local solutions to global ones, provided a uniform bound exists.

• Integral Group-Algebra and Type-Hecke Conditions (Section 6):
  The paper formulates necessary and sufficient conditions for the existence of invariant norms in terms of:

  1. Emerton's Condition (EC): A spectral condition on the Jacquet modules ensuring eigenvalues of positive Jacquet operators have absolute value $\le 1$.
  2. Finite Type–Bruhat Quotients: The transition graphs of matrix coefficients must descend to a finite graph where every directed cycle has non-negative cost.
  3. Integral Type Intertwiners: The matrices relating different types must be integral.

• Explicit Analysis for $G = \text{GL}_n(F)$ (Section 7 & 8):
  For $G = \text{GL}_n(F)$ and $V$ associated with a potentially semistable representation, the author provides an explicit combinatorial framework:

  • Multisegment Reduction: The problem is translated into the language of Bernstein–Zelevinsky multisegments and affine Hecke algebras.
  • Global Reduction Matrix: A "global reduction matrix" $R$ is constructed via a deterministic pivot rule on an affine–Bruhat shelling. This matrix expresses the action of group elements in a normal-form basis.
  • Path Formula: Matrix coefficients are expressed as sums over paths in a reduction graph. The valuation of these coefficients is bounded below by a sum of "defects" associated with admissible Weil–Deligne prefixes.
  • Sufficient Conditions (Proposition 8.8): For a single segment, the paper proves that if the defects $\Delta_k$ are non-negative (admissibility), the type intertwiners are integral, and the reduction paths are "positively oriented" (noncentral translation content lies in the dominant cone), then the standard max-norm is invariant under the Cartan generators.

4. Specific Findings on $G = \text{GL}_n(F)$
In the case of a single segment (generalized Steinberg representations), the paper demonstrates that the existence of an invariant norm follows if:

1. The filtered Weil–Deligne module is admissible (implying non-negative defects $\Delta_k \ge 0$).
2. The central character is unitary.
3. The reduction paths satisfy a positivity condition regarding their noncentral translation content.
   The author explicitly verifies these conditions for the block Cartan directions using Haines–Pettet formulas for minuscule coweights, showing that the "positive-prefix" hypothesis holds for these specific generators.

5. Significance and Claims
The paper claims to provide a direct geometric proof that boundedness of an orbit implies the existence of an invariant norm, bypassing the need for CAT(0) circumcenter arguments which do not apply to the $\ell_\infty$ metric.

The primary significance lies in isolating the geometric mechanism from the arithmetic difficulty. The author states that while the geometric fixed-point step is elementary and complete, the substantive arithmetic challenge remains: constructing a norm whose entire orbit is uniformly bounded. This requires verifying the "finite-state" conditions (non-negative cycle costs in the transition graph) which depend on deep representation-theoretic properties (integrality of type intertwiners and positivity of defects).

The paper does not claim to solve the Breuil–Schneider conjecture in full generality but provides the necessary framework and explicit criteria (Proposition 7.22, Proposition 8.8) to reduce the problem to checking specific combinatorial and valuation-theoretic inequalities within the affine-Weyl group and type modules. It emphasizes that for $G = \text{GL}_n(F)$, the problem reduces to verifying that the "defects" in the multisegment reduction are non-negative and that the reduction paths respect the dominant cone.