OctoInformBeta

OpenAI preprint, September 2026 · §5. From an ultrapower embedding to a finite contradiction

Lemma 5.1: For unitary families that become unital and multiplicative in operator norm

Ultrafilter limits retain the Kazhdan gap.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Unitary families may act on different Hilbert spaces at each index, with their multiplication laws holding only asymptotically in operator norm. Their least quadratic displacement energy on a fixed finite subset of the group is a scalar quantity that can be compared across these spaces.

Let Γ\Gamma be a group, let S⊆ΓS\subseteq\Gamma be finite, and put R=4|S|R=4|S|. Let ω\omega be an ultrafilter on an index set II, and let EiE_i be complex Hilbert spaces, with inner products linear in the second variable. For each i∈Ii\in I and g∈Γg\in\Gamma, let Ui(g)U_i(g) be a unitary operator on EiE_i.

The quadratic displacement energy Qi(ξ)Q_i(\xi) of a vector ξ∈Ei\xi\in E_i is the nonnegative real number given by the sum in (1): Qi(ξ)=∑s∈S∥Ui(s)ξ−ξ∥2.Q_i(\xi)=\sum_{s\in S}\|U_i(s)\xi-\xi\|^2.(1)✦ The bottom Rayleigh energy rir_i of UiU_i over SS is the real infimum of its attainable unit-vector energies, as in (2), with value zero when Ei={0}E_i=\{0\}: ri=inf⁡{Qi(ξ):ξ∈Ei,∥ξ∥=1}.r_i=\inf\bigl\{Q_i(\xi):\xi\in E_i,\ \|\xi\|=1\bigr\}.(2)✦

A pair (S,κ)(S,\kappa) consisting of a finite subset SS of a group Γ\Gamma and a real number κ\kappa satisfies the Kazhdan gap inequality if every unitary representation ρ\rho of Γ\Gamma on a complex Hilbert space HH whose only invariant vector is zero satisfies (3): κ∥ξ∥2≤∑s∈S∥ρ(s)ξ−ξ∥2(ξ∈H).\kappa\|\xi\|^2 \leq \sum_{s\in S}\|\rho(s)\xi-\xi\|^2 \qquad (\xi\in H).(3)✦ A family UiU_i of unitary operators Ui(g)U_i(g) on complex Hilbert spaces EiE_i, indexed by i∈Ii\in I and g∈Γg\in\Gamma, is asymptotically unital and multiplicative along an ultrafilter ω\omega on II if its operator norm defects satisfy (4) for every g,h∈Γg,h\in\Gamma: limi→ω∥Ui(1)−1Ei∥=0,limi→ω∥Ui(gh)−Ui(g)Ui(h)∥=0.\lim_{i\to\omega}\|U_i(1)-1_{E_i}\|=0, \qquad \lim_{i\to\omega}\|U_i(gh)-U_i(g)U_i(h)\|=0.(4)

For a unitary tuple xnx_n and a unitary representation πn\pi_n, the paper’s notation d(xn,πn)d(x_n,\pi_n) denotes the bottom energy of the tensor tuple with second factor the conjugate representation πn¯\overline{\pi_n}. The library uses πn\pi_n directly in the second tensor factor.

The Claim

Theorem 1. Let UiU_i be a unitary family as above. Its bottom energies converge to a real number rωr_\omega as follows: ri→rω(i→ω).r_i\longrightarrow r_\omega\qquad(i\to\omega).(5)✦ Suppose that (S,κ)(S,\kappa) satisfies (3) and that UiU_i satisfies (4). The limit in (5) then satisfies rω∈{0}∪[κ,R].r_\omega\in\{0\}\cup[\kappa,R].(6)✦ Suppose further that I=ℕI=\mathbb N. For every real a<κa<\kappa such that {i:ri<a}∈ω\{i:r_i<a\}\in\omega, the bottom energies satisfy ri→0(i→ω).r_i\longrightarrow0\qquad(i\to\omega).(7)✦

In the paper, the result reads: for every sequence πn\pi_n of nonzero finite-dimensional unitary representations of Γ\Gamma, with no bound on their dimensions, lim⁡ωd(xn,πn)\lim_\omega d(x_n,\pi_n) belongs to {0}∪[κ,R]\{0\}\cup[\kappa,R], and if {n:d(xn,πn)<a}∈ω\{n:d(x_n,\pi_n)<a\}\in\omega with a<κa<\kappa, that limit is zero.

What Is Checked

The convergence assertion (5) is stated over arbitrary unitary families on varying complex Hilbert spaces; the gap assertions (6) and (7) additionally assume the Kazhdan inequality and operator norm convergence of the identity and multiplication defects.