REVIEW 2 major objections 5 minor 2 references
Energy, Bosons and Computational Complexity
T0 review · 2 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash
Pith's one-line read Average photon number is a computational resource: with exponential energy, bosonic circuits solve NP; with unbounded energy, they reach PTOWER.
desk verdict The lower-bound half of this paper is genuinely new and valuable; the PP upper-bound proof has a fixable but real squeezing-budget gap. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing construction is the time-independent Hamiltonian H = P̂₀ + A(X̂₀)N̂₄ₖ₊₃ of Eq. (88), which encodes a time-dependent adiabatic evolution A(t) using a continuous position clock and a high-energy register that rescales time to constant physical duration. Its logical restriction A(t) preserves photon number and has a unique ground state with spectral gap Ω(n_max^{-2}) by mapping the problem to the Laplacian of a 'whiskered grid graph' (Lemma 4.7). Other essential pieces: the photon-number-controlled squeezer that doubles energy per application and produces the astronomical input energies needed for PTOWER; the detuned degenerate parametric amplifier gadget (Hamiltonian G(c), Eq.
What would settle it
Compute the low-lying spectrum of the full unbounded Hamiltonian of Eq. (88) on a sequence of increasing photon-number truncations: if the gap between the two lowest eigenvalues shrinks to zero (or many eigenvalues intrude below the claimed bound) as the cutoff grows, Lemma 4.7's gap analysis fails in the infinite-dimensional limit and the PTOWER containment collapses. Alternatively, exhibit a CV circuit with energy o(ε^{-1}) that solves the ε-BeamSplitPrec problem, contradicting the energy hierarchy theorem.
Extended reading notes
Core claim
The central discovery is that the energy of a bosonic computation, measured as the expected photon number ⟨ψ|N|ψ⟩, is a genuinely computational resource with a strict hierarchy: more energy buys strictly more power. Concretely, the paper constructs finite gate sets for which polynomial-time CVBQP with exponential energy and O(1) modes solves NP (Theorem 4.1), with O(k) modes solves NTIME(exp^(k)), and with no energy bound contains PTOWER (Theorems 4.2 and 4.3). The proof goes through a CV adiabatic algorithm for Diophantine equations whose evolution time is made constant by a photon-number-preserving Hamiltonian with a provably polynomial spectral gap over a 'whiskered grid graph'; the high-
Load-bearing premise
The NP and PTOWER lower bounds hinge on the assumption that the spectral gap and unique ground state proven for the adiabatic Hamiltonian on a finite-dimensional logical subspace survive when the same Hamiltonian is embedded in the full infinite-dimensional operator of Eq. (88), whose essential self-adjointness is only sketched in the paper.
Editorial extensions
If this is right
- Exponential energy with O(1) modes suffices for CVBQP to contain NP, giving formal evidence that constant-mode, exponential-energy setups (such as recent CV factoring proposals) are computationally very strong.
- Without energy restrictions, CVBQP with a suitable finite gate set contains PTOWER; any efficient simulation of these circuits by Gaussian-and-cubic circuits would violate the time hierarchy theorem.
- Polynomial-energy CV computations are simulable by BQP/poly, and by BQP for gate sets with efficient block encodings (including Kerr, Gaussians, Jaynes–Cumming, and cubic phase), so low energy does not grant quantum advantage beyond BQP.
- The Gaussian-plus-cubic gate set with exponential energy lies in PP, improving the known PSPACE bound and placing a common non-Gaussian gate set 'not too far' from BQP.
- Finite-energy but otherwise unconstrained 'physical' CVBQP computations are decidable (in R); undecidability sets in only with unbounded or infinite-energy states.
Reading between the lines
- If the spectral-gap/unbounded-operator analysis is later completed, the adiabatic Diophantine construction would become the first rigorous CV adiabatic algorithm for NP/PTOWER with bounded runtime; until then, the lower bounds rest on an infinite-dimensional gap assumption that needs independent checking.
- The PP upper bound for Gaussian+cubic suggests any quantum advantage from this gate set with exponential energy would not reach beyond the counting hierarchy; a natural next step is to test whether the bound can be tightened to BQP or whether a matching NP-hardness lower bound can be shown for the same gate set.
- The energy hierarchy theorem (ε-BeamSplitPrec) offers a concrete experimental probe: with energy budget E=o(ε^{-1}) a single query should fail to detect the beamsplitter angle, while E=O(ε^{-2}) should succeed — a distinction testable in photonic platforms.
- The infinite-energy-in-finite-time results imply that general polynomial-Hamiltonian gate sets are not physically realisable to arbitrary precision, so experimental roadmaps should restrict to energy-constrained or block-encoded gate sets to avoid unphysical regimes.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies energy (average photon number) as a computational resource in continuous-variable bosonic circuits. It proves three groups of results: (1) quantitative energy-growth bounds for specific gate sets, including doubly exponential growth for Gaussian+cubic-phase gates and infinite-energy states in constant time; (2) complexity lower bounds, showing that certain finite gate sets with exponential energy and O(1) modes contain NP, and with more modes/energy contain ELEMENTARY and even PTOWER; (3) simulation upper bounds, including CVBQP with polynomial energy in BQP/poly, decidability of physical CV computations with finite energy, and containment in PP of Gaussian+cubic-phase circuits with exponential energy. The paper further combines the upper and lower bounds to argue against efficient continuous-variable Solovay–Kitaev theorems for the Gaussian+cubic gate set.
Significance. If the main theorems are fully established, this is a substantial contribution: it makes precise how energy can act as a computational resource in infinite-dimensional bosonic systems, connects to the BCCK factoring setup, gives new upper bounds (PP) for a physically relevant gate set, and provides a novel no-go argument for CV Solovay–Kitaev. The energy-growth calculations in Section 3 are self-contained and checkable, and the BQP/poly simulation of Section 5.1 is a useful general tool. The paper is also honest in flagging some of its own limitations, notably the deferred unbounded-operator analysis in Appendix A and the conditional nature of the exponential-energy promise in the PP result.
major comments (2)
- [§4.1.5–4.1.9, Lemmas 4.7–4.15, Appendix A] The NP/PTOWER lower bounds rest on spectral-gap and unbounded-operator claims that are not fully proved. Lemma 4.7 establishes the gap Ω(n_max^{-2}) only for A(t) restricted to a single logical subspace Hlog_n; the full Hamiltonian in Eq. (88) is unbounded and acts on infinitely many photon-number invariant sectors, and the paper does not give a uniform gap bound or a rigorous adiabatic theorem for this unbounded operator. Lemma 4.10's essential self-adjointness argument is a sketch ('It suffices to check...'), and Appendix A explicitly says a more complete unbounded-operator treatment is planned for a future revision. Since Lemma 4.15 and the E_T fast-forwarding are load-bearing for Theorems 4.1–4.3, these containments are conditional on completing that analysis.
- [§5.5.1, proof of Theorem 5.27, Eqs. (314), (322)–(327)] The inductive invariant that all coefficients and Gaussian parameters remain ≤exp(n) under adaptive homodyne postselection is underproved. The update rule for the mean vector in Eq. (327) is asserted without derivation, and the coefficient update in Eq. (322) is only bounded by combining Lemma 5.32 with the normalization product in Eq. (323). A single teleportation step can multiply a coefficient by exp(n)/√Z ≈ exp(n/2) if ξ=exp(n), so after T=poly(n) cubic gates the coefficients can grow to exp(poly(n)), not necessarily ≤exp(n). The text may intend a global rescaling, but this is not stated or proved. Since the PP simulation requires each branch to be generated and evaluated in polynomial time, this invariant needs a complete proof or a revised parametrization.
minor comments (5)
- [Appendix A] The sentence 'We plan to include a more complete overview over the required theory of unbounded operators in a future revision' is inappropriate for a submitted manuscript if it covers load-bearing technical material. Either complete the treatment or explicitly mark the affected theorems as conditional.
- [Theorem 5.27] The phrase 'using one query to a PP oracle' is redundant since P^{PP[1]}=PP; the statement should simply say the problem is in PP.
- [Eq. (315)] The notation (^n⊗I) appears to denote the Fock-state projector |n⟩⟨n|⊗I, not the number operator; this should be clarified.
- [Theorem 5.20 and Lemma 5.9] The symbol ε is used both for the teleportation error in Lemma 5.9 and for the width parameter in the Gaussian decomposition of Theorem 5.20. These are different parameters and should be renamed to avoid confusion.
- [Lemma 5.21] Typo: 'Reimann' should be 'Riemann'.
Circularity Check
No significant circularity: lower and upper bounds are new constructions, not re-definitions of their inputs.
full rationale
The paper's derivation chain is not circular. The CV model is inherited from prior work [CJMM25], but none of the advertised containments (NP⊆CVBQP[G], PTOWER⊆CVBQP[G], CVBQP_poly⊆BQP/poly, CVBQP[X^3]_exp⊆PP) is obtained by fitting parameters to the target result or by assuming the conclusion. The NP/PTOWER lower bounds are built from a novel time-independent adiabatic Hamiltonian with an explicit spectral-gap analysis (Lemma 4.7) and error bounds (Lemma 4.14), and the PP upper bound uses an explicit teleportation gadget plus a Gaussian-rank decomposition of cubic phase states (Theorem 5.20). Self-citations to [CJMM25], [URC25], and [ABC25] supply background lemmas and technical ingredients, but they are not uniqueness theorems invoked to forbid alternatives, and the central results do not reduce to those citations. The skeptic-flagged mismatch between the squeezing requirement in Lemma 5.9 (ξ>c(E/ε)exp(½log²(2/δ)log(1/δ))) and the ξ=exp(n) regime used in Corollary 5.30 is a quantitative proof-consistency issue, not a circularity: it concerns whether the stated error bound is achieved with the stated parameters, not whether the output was assumed as input. Likewise, Appendix A's statement that a fuller unbounded-operator treatment is planned for a future revision flags a rigor gap in Lemma 4.10, but an omitted proof is not an equivalence-by-construction. Therefore no circular step is exhibited, and the appropriate circularity score is 0.
Assumptions & free parameters
free parameters (3)
- Squeezing parameter r in lower-bound construction =
poly(n)
- Adiabatic time-register energy E_T =
sufficiently large, E^{O(k)}
- Teleportation squeezing ξ in Theorem 5.27 =
≤ exp(n)
assumptions (8)
- standard math Matiyasevich–Robinson–Davis–Putnam theorem: Diophantine equations are undecidable
- standard math Adleman–Manders NP-completeness of binary quadratic Diophantine equations and solution bounds
- standard math Adiabatic theorem with error bounds (Reichardt form)
- standard math Spectral gap lower bound for whiskered grid graphs (BPS07 formula)
- standard math Time hierarchy theorem and PTOWER ⊄ EXPSPACE
- standard math Essential self-adjointness criteria (Stone, Nelson, Kato–Rellich)
- domain assumption Assumption 5.1: polynomial-time block encodings for the gate set at bounded energy
- domain assumption Proxy-state promise in Theorem 5.27: a state of energy exp(n) exists that is exponentially close to the true state
Cite this review
Pith. "Pith review of Energy, Bosons and Computational Complexity." pith.science (2026). https://pith.science/paper/6NJGUK3R
@misc{pith2026251008545,
author = {Pith},
title = {Pith review of: Energy, Bosons and Computational Complexity},
year = {2026},
howpublished = {\url{https://pith.science/paper/6NJGUK3R}},
note = {Machine review of arXiv:2510.08545}
}
abstract
We investigate the role of energy, i.e. average photon number, as a resource in the computational complexity of bosonic systems. We show three sets of results: (1. Energy growth rates) There exist bosonic gate sets which increase energy incredibly rapidly, obtaining e.g. infinite energy in finite/constant time. We prove these high energies can make computing properties of bosonic computations, such as deciding whether a given computation will attain infinite energy, extremely difficult, formally undecidable. (2. Lower bounds on computational power) More energy ``='' more computational power. For example, certain gate sets allow poly-time bosonic computations to simulate PTOWER, the set of deterministic computations whose runtime scales as a tower of exponentials with polynomial height. Even just exponential energy and $O(1)$ modes suffice to simulate NP, which, importantly, is a setup similar to that of the recent bosonic factoring algorithm of [Brenner, Caha, Coiteux-Roy and Koenig (2024)]. For simpler gate sets, we show an energy hierarchy theorem. (3. Upper bounds on computational power) Bosonic computations with polynomial energy can be simulated in BQP, ``physical'' bosonic computations with arbitrary finite energy are decidable, and the gate set consisting of Gaussian gates and the cubic phase gate can be simulated in PP, with exponential bound on energy, improving upon the previous PSPACE upper bound. Finally, combining upper and lower bounds yields no-go theorems for a continuous-variable Solovay--Kitaev theorem for gate sets such as the Gaussian and cubic phase gates.
Figures
Reference graph
Works this paper leans on
-
[2005]
A Scheme for Efficient Quantum Computation with Linear Optics
arXiv:quant-ph/0504101 [quant-ph]. [Kie06] T. D. Kieu.On the identification of the ground state based on occupation probabilities: An investigation of Smith’s apparent counterexamples. 2006. arXiv:quant-ph/0602145 [quant-ph]. [KLM01] E. Knill, R. Laflamme, and G. Milburn. “A Scheme for Efficient Quantum Computation with Linear Optics”. In:Nature409 (2001)...
arXiv 2006
-
[2019]
New York, NY, USA: Association for Computing Machinery, 2019, pp. 193–204. doi:10.1145/3313276.3316366. [Hal13] B. C. Hall.Quantum Theory for Mathematicians. Springer New York, 2013.doi:10 . 1007/978-1-4614-7116-5. [HJ12] R. A. Horn and C. R. Johnson.Matrix Analysis. 2nd ed. Cambridge University Press, 2012. 71 [HKS+17] C. S. Hamilton, R. Kruse, L. Sanson...
arXiv 2019
Reviewed August 4, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.