The minimal completion for the spectral target
Curvature is insufficient for Qspec. Holonomy is manifestly sufficient, since by Proposition 10.1 the spectrum is a function of θ mod 2π. The question this section answers is the sharp one: how much less than holonomy is enough? The answer is: strictly less. The spectral target sees holonomy only up to inversion.
Holonomy is sufficient
Proposition 11.1 (Holonomy completes the curvature readout for the spectrum).
The identity map e=id:Sflat→Sflat, equivalently the holonomy map h:Sflat→U(1), h(θ)=eiθ, is a readout extension of pF that is sufficient for Qspec.
Proof. h is a bijection Sflat→U(1) by (32) and (33), hence surjective, and pF=u∘h with u:U(1)→{0} the constant map, so h is a readout extension in the sense of Definition 6.1. Define b:U(1)→Aspec by b(eiθ):=Spec(Hθ); this is well defined because, as shown at the end of the proof of Proposition 10.1, Spec(Hθ) depends only on θ mod 2π, i.e. only on eiθ. Then aspec=b∘h, so h is Qspec-sufficient by Definition 6.3. ◻
Spectral rigidity: the exact equivalence
Theorem 11.2 (Isospectrality iff θ′≡±θ).
For θ,θ′∈R, Spec(Hθ)=Spec(Hθ′)⟺θ′≡θ or θ′≡−θ(mod2π),(48) where the spectra are compared as multisets, with multiplicities.
Proof. (⇐) Reverse direction. Suppose first θ′≡θ(mod2π). Then Spec(Hθ′)=Spec(Hθ) by the invariance established at the end of the proof of Proposition 10.1.
Suppose next θ′≡−θ(mod2π). By the same invariance it suffices to treat θ′=−θ. Then for every n∈Z, λ−n(−θ)=(L2π(−n)−(−θ))2=(L−(2πn−θ))2=(L2πn−θ)2=λn(θ).(49) The map n↦−n is a bijection of Z onto itself. Hence (49) exhibits a bijection between the indexing set of Spec(H−θ) and that of Spec(Hθ) carrying each eigenvalue to an equal eigenvalue. A multiset is determined by its indexed family up to bijective reindexing, so the two multisets are equal, with all multiplicities: m−θ(λ)=#{n:λn(−θ)=λ}=#{n:λ−n(−θ)=λ}=#{n:λn(θ)=λ}=mθ(λ). So Spec(H−θ)=Spec(Hθ). Concretely, the reindexing n↦−n is implemented on the Hilbert space by complex conjugation, ψn↦ψn=ψ−n, which is an antiunitary intertwiner between Hθ and H−θ; but the elementary reindexing argument above is already complete and requires nothing further.
(⇒) Forward direction. Suppose Spec(Hθ)=Spec(Hθ′). We recover θ up to sign and 2π from the spectrum, using the lowest eigenvalue.
Define, for any θ∈R, the principal representative α(θ)∈[0,π]the unique number in [0,π] with θ≡α(θ) or θ≡−α(θ)(mod2π).(50) Existence and uniqueness of α(θ): reduce θ mod 2π to a representative θ~∈[0,2π). If θ~∈[0,π] set α:=θ~; if θ~∈(π,2π) set α:=2π−θ~∈(0,π), and then θ≡θ~=−α(mod2π). In both cases α∈[0,π] and θ≡±α. For uniqueness, suppose α,β∈[0,π] both work; then α≡±β(mod2π). If α≡β then α−β∈2πZ and ∣α−β∣≤π<2π, forcing α=β. If α≡−β then α+β∈2πZ with 0≤α+β≤2π, so α+β∈{0,2π}, giving either α=β=0 or α=β=π. In all cases α=β.
We claim λmin(θ):=minSpec(Hθ)=L2α(θ)2.(51) To see this, note λn(θ)=(L2πn−θ)2, so n∈Zminλn(θ)=L21n∈Zmin(2πn−θ)2=L21(dist(θ,2πZ))2, the minimum being attained since (2πn−θ)2→∞ as ∣n∣→∞. It remains to identify dist(θ,2πZ)=α(θ). Write θ≡±α with α=α(θ)∈[0,π]; since 2πZ is invariant under negation and under translation by 2π, we get dist(θ,2πZ)=dist(±α,2πZ)=dist(α,2πZ). For α∈[0,π] the nearest point of 2πZ is 0, since the next candidates are ±2π at distance 2π−α≥π≥α and 2π+α>α. Hence dist(α,2πZ)=α, proving (51).
Now assume Spec(Hθ)=Spec(Hθ′). Equal multisets have equal minima, so by (51) L2α(θ)2=λmin(θ)=λmin(θ′)=L2α(θ′)2, whence α(θ)2=α(θ′)2. Since α(θ),α(θ′)∈[0,π] are both nonnegative, α(θ)=α(θ′)=:α. By (50), θ≡±α and θ′≡±α(mod2π). Combining the four sign cases: if θ≡α and θ′≡α then θ′≡θ; if θ≡α and θ′≡−α then θ′≡−θ; if θ≡−α and θ′≡α then θ′≡−θ; if θ≡−α and θ′≡−α then θ′≡θ. In every case θ′≡±θ(mod2π), as required. ◻
Corollary 11.3 (The spectral target sees holonomy only modulo inversion).
Define the equivalence ∼spec on Sflat≅U(1) by eiθ∼speceiθ′⟺θ′≡±θ(mod2π)⟺eiθ′∈{eiθ,e−iθ}.(52) Then ∼spec is exactly the answer equivalence of the spectral target: aspec(θ)=aspec(θ′) iff θ∼specθ′.
The coarsest sufficient completion and its coordinate
Theorem 11.4 (cosθ coordinatizes the minimal spectral completion).
Let c:Sflat⟶[−1,1],c(θ):=cosθ.(53) Then:
c is a well-defined surjection onto [−1,1];
c(θ)=c(θ′) if and only if θ∼specθ′; that is, c realizes exactly the spectral equivalence classes;
c is a readout extension of pF that is sufficient for Qspec;
c is a coarsest such extension: the induced map φ[−1,1]:[−1,1]→EQspec of Theorem 8.3 is a bijection. Consequently every Qspec-sufficient readout extension e must separate source states with distinct values of cosθ.
Proof. (1) cos is 2π-periodic, so c descends to R/2πZ; its image is [−1,1] since cos restricted to [0,π] is already onto [−1,1].
(2) The function cos is strictly decreasing on [0,π], hence injective there. Given θ,θ′, write α=α(θ),α′=α(θ′)∈[0,π] as in (50); since cos is even and 2π-periodic, cosθ=cosα and cosθ′=cosα′. Therefore c(θ)=c(θ′)⟺cosα=cosα′⟺α=α′⟺θ′≡±θ(mod2π)⟺θ∼specθ′, where the second equivalence uses injectivity of cos on [0,π], and the third is the argument at the end of the proof of Theorem 11.2 (equality of principal representatives is equivalent to θ′≡±θ).
(3) c is surjective by (1), and pF=u∘c with u:[−1,1]→{0} constant, so c is a readout extension. For sufficiency, define bspec:[−1,1]→Aspec by bspec(κ):=Spec(Harccosκ),arccosκ∈[0,π].(54) If c(θ)=κ then by (2) θ∼specarccosκ (both have principal representative arccosκ), so by Corollary 11.3 Spec(Hθ)=Spec(Harccosκ)=bspec(c(θ)). Hence aspec=bspec∘c, and c is Qspec-sufficient.
(4) Write φ:=φ[−1,1] and Espec:=EQspec for brevity. By Theorem 8.3, applied to the sufficient extension c, there is a unique surjection φ:[−1,1]→Espec with jQspec=φ∘c, given by φ(κ)=(u(κ),bspec(κ))=(0,bspec(κ)),κ∈[−1,1].(55) It remains to prove injectivity. Suppose φ(κ)=φ(κ′). Comparing second coordinates in (55) gives bspec(κ)=bspec(κ′), which by (54) says Spec(Harccosκ)=Spec(Harccosκ′).(56) By Theorem 11.2, (56) forces arccosκ′≡±arccosκ(mod2π). Both arccosκ and arccosκ′ lie in [0,π], so each is its own principal representative in the sense of (50); by the uniqueness established there, they are equal. Applying cos yields κ=κ′.
Hence φ is a bijection, and c carries precisely the distinctions of Espec , no more and no fewer.
The final assertion follows from Corollary 8.4: if e is Qspec-sufficient and c(θ0)=c(θ1), then by (2) θ0∼specθ1, so by Corollary 11.3 aspec(θ0)=aspec(θ1), and Theorem 7.1 forces e(θ0)=e(θ1). ◻
Example 11.6 (A target for which cosθ is insufficient: oriented transport).
Consider the target QHol with AQHol=U(1) and correct-answer map aHol(θ)=eiθ: “what phase does a unit charge acquire on being transported once around the circle in the positive direction?” Take θ=π/2 and θ′=−π/2. Then c(π/2)=cos(π/2)=0=cos(−π/2)=c(−π/2), so c does not separate these two states; but aHol(π/2)=eiπ/2=i=−i=e−iπ/2=aHol(−π/2). By Theorem 7.1, c is not QHol-sufficient. By Theorem 11.2 the two states are nonetheless isospectral, Spec(Hπ/2)=Spec(H−π/2), consistent with c being sufficient for Qspec.
The full holonomy map h is QHol-sufficient (take b=idU(1)), and by Corollary 8.4 it is coarsest for that target, since jQHol=(pF,aHol) has the same level sets as h. So on one and the same reduced source space, with one and the same readout, the minimal completion for the transport target is the full holonomy circle, while the minimal completion for the spectral target is its strictly coarser inversion quotient U(1)/(z∼zˉ)≅[−1,1]. The scalar spectrum is blind to orientation; oriented transport is not. This is target-relativity of completion in its sharpest form, and it is not a defect of either description: each retains exactly what its question requires.