{"id":"e13c7307-5928-4219-95fa-e4027780a48a","arxiv_id":"1909.00028","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A discontinuous Galerkin basis built by blockwise SVD compression of primitive diagonal bases yields block-diagonal two-electron integrals and a predicted quantum-simulation cost crossover at 15 to 20 atoms for hydrogen chains.","lead":"This paper introduces a new way to discretize molecules for quantum chemistry that keeps the Hamiltonian sparse and block-diagonal, reducing the predicted quantum simulation cost for hydrogen chains from O(N^4.5) to O(N^2.6). The method also speeds up classical DMRG calculations by one to two orders of magnitude while preserving accuracy, with the crossover around 15 to 20 atoms.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(N^2) integral scaling and O(N^2.6) quantum-cost claim depend on n_kappa staying constant at fixed accuracy; this is supported only by Figure 8 at fixed SVD tolerance, and the paper's own tighter-tolerance fits show the scaling degrading.","rationale":"I read the paper as a methodological proposal, and the core construction (blockwise SVD of the primitive-to-active-space projection to obtain a block-diagonal two-electron tensor) is sound; the numerical demonstrations on hydrogen chains are credible and no red flags indicate misconduct. The load-bearing concern is exactly the one the reader identified: the asymptotic scaling rests on n_kappa remaining constant for fixed accuracy, a premise that is asserted in Section III and supported only by Figure 8 at fixed SVD thresholds. My refinement is that fixed tau is not the same as fixed physical accuracy, and the paper's own data show the fitted exponents drift upward as tau tightens, so the advertised O(N^2.6) is not obviously robust. This does not change the verdict: the paper should remain CONDITIONAL, with the condition being an explicit test of the n_kappa-constant premise and the fixed-accuracy scaling behavior.","tokens_in":26742,"tokens_out":9160,"duration_ms":91254,"concrete_test":"Re-run the DG construction for H chains N=10,20,30,40 (and, if feasible, a 2D or 3D periodic or strongly correlated test) with the same primitive and active spaces, but instead of fixing tau, increase n_kappa per block until the DG-basis energy per atom is within a fixed tolerance (e.g., 1 mHa) of the exact cc-pVDZ active-space energy, and report n_kappa(N) at that fixed accuracy. If n_kappa grows with N, or if the exponent alpha_L at fixed per-atom accuracy exceeds the values in Figure 9, the O(N^2) integral scaling and the O(N^2.6) quantum cost claim are not asymptotic in N.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is the asymptotic scaling reduction from O(N^4.5) to O(N^2.6) for fault-tolerant evolution, which follows from L = O(N_b^2 n_kappa^4) and lambda = O(N^1.5) only if n_kappa remains O(1) as system size grows. Section III states this as an expectation (\"we expect ... n_kappa ... bounded by a constant\") and Figure 8 is offered as substantiation. The weakness is that Figure 8 fixes the SVD tolerance tau, not the physical accuracy, and no error analysis shows that a fixed tau keeps the per-atom (or total) energy error bounded as N grows. If the active space contains delocalized or strongly correlated orbitals, the number of significant singular values per spatial block can grow with N; the fits in Figure 9 are sensitive to tau (alpha_L = 2.03, 2.18, 2.34 for tau = 1e-1, 1e-2, 1e-3; alpha_lambda = 1.42, 1.47, 1.58), giving cost exponents 2.44-2.75 rather than a robust 2.6. Since the crossover at 15-20 atoms and the claimed advantage both depend on these fitted exponents, the headline scaling is not established for fixed accuracy in general systems.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a discontinuous Galerkin (DG) discretization for second-quantized electronic structure, obtained by partitioning a diagonal primitive basis into spatial blocks and SVD-compressing the projection of an active-space basis onto each block. The resulting DG basis has a block-diagonal two-electron tensor, interpolating between dense molecular-orbital bases with O(N^4) integrals and diagonal primitive bases with O(N^2) integrals. For quantum simulation, the authors design swap networks and LCU-based methods exploiting the block structure, and they report empirical scaling for hydrogen chains: the number of nonzero two-electron integrals grows as roughly N^{2.0-2.3}, lambda as N^{1.4-1.6}, leading to a claimed fault-tolerant evolution cost of O(N^{2.6}) versus O(N^{4.5}) for the Gaussian active-space basis, with a crossover around 15-20 atoms. They also test the DG basis in classical DMRG calculations, reporting one to two orders of magnitude speedups while maintaining accuracy relative to the complete basis set limit.","tokens_in":27061,"tokens_out":5108,"duration_ms":46306,"significance":"If the central scaling claims hold, this is a valuable contribution: it provides a concrete, systematic way to interpolate between diagonal and compact non-diagonal discretizations, with algorithmic machinery (swap networks, LCU preparation, and a hybrid active-space construction) that is likely to be useful beyond the specific hydrogen-chain tests. The block-diagonal sparsity pattern is derived cleanly from the SVD construction and is internally consistent. The paper also makes an honest empirical contribution by testing the representation in correlated DMRG calculations and reporting crossover data, rather than only presenting asymptotic arguments. However, the headline quantitative claims rest on fitted exponents and on the unproven constancy of the per-block DG basis count n_kappa at fixed physical accuracy; these are empirical rather than derived, and the manuscript itself labels the constancy of n_kappa as an expectation rather than a proven property.","major_comments":[{"comment":"The asymptotic claims O(N_d^2) integral scaling, O(N^{2.6}) fault-tolerant evolution cost, and the 15-20 atom crossover all depend on n_kappa remaining O(1) at fixed accuracy as the system grows. Section III states this as an expectation (\"we expect ... n_kappa ... bounded by a constant\"), and Figure 8 substantiates it only for fixed SVD truncation tolerance tau, not for a fixed physical accuracy target. No convergence analysis links tau to the per-electron or total energy error, so the data do not exclude n_kappa growing with system size for delocalized, metallic, or strongly correlated systems. If n_kappa grows, the claimed scaling advantages degrade. The manuscript should either provide an error bound or rigorous argument for n_kappa at fixed accuracy, or present additional numerical evidence with error bars, including a test case with delocalized or strongly correlated orbitals.","section":"Section III, Eq. (8), and Figure 8"},{"comment":"The headline exponent O(N^{2.6}) is obtained by fitting log-log slopes to L and lambda over hydrogen chains from N=2 to N=32 with no reported residuals or confidence intervals and with an integral-counting threshold of 10^{-6}. The fitted exponents themselves vary with the SVD tolerance (alpha_L = 2.03, 2.18, 2.34 and alpha_lambda = 1.42, 1.47, 1.58 for tau = 10^{-1}, 10^{-2}, 10^{-3}), giving cost exponents between about 2.44 and 2.75 rather than a robust 2.6. The crossover is also parameter-dependent, ranging from H6-H8 to H20-H22 in Table I depending on tolerance and bond length. The authors should report the fitting procedure and uncertainties, quantify sensitivity to the integral-counting threshold, and present the crossover as a range of system sizes for specific tolerances rather than as a single universal statement.","section":"Section V A, Figures 9-10, Table I"},{"comment":"The comparison baseline O(N^{4.5}) is described as applying \"when not exploiting further low rank structure,\" yet the text immediately notes that state-of-the-art molecular-orbital algorithms use low-rank factorization to achieve O(N_a^{3/2} lambda t). With the paper's own empirical lambda ~ N^{2.5}, that improved molecular-orbital cost would be O(N^4), not O(N^{4.5}). Since the abstract and conclusion advertise a reduction \"from O(N^{4.5}) to O(N^{2.6}),\" the baseline should be clarified and benchmarked against the best available molecular-orbital algorithm, or the authors should explicitly justify why the low-rank improvement is unavailable in the comparison. The qualitative DG advantage may survive this correction, but the headline factor would change.","section":"Section V A and Conclusion"}],"minor_comments":[{"comment":"The text says \"the average number of DG-basis per atom\" while the construction groups primitive functions into spatial blocks, and in these hydrogen-chain runs there is one block per atom; please clarify the relation between blocks and atoms consistently in text, captions, and axis labels.","section":"Section V A, Figure 8"},{"comment":"The notation \"DG 10 1\" should read \"DG 10^{-1}\" or similar, and the exponents in the legends appear without their defining symbol; please label them as alpha or state explicitly that the shown numbers are the fitted power-law exponents.","section":"Figures 9-10 and Table I"},{"comment":"The notation excludes terms \"{p = r, q = s}\" but the meaning of p, q, r, s here is not defined in the appendix; please spell out the index convention used in the lambda computation.","section":"Appendix B, Eq. (B1)"},{"comment":"The sentence giving the depth as O(N_b n_kappa^3) = O(N_d n_kappa^2) is correct only up to constants and the assumption of roughly equal block sizes; it would help to state that this is an asymptotic statement under that assumption.","section":"Section IV A"},{"comment":"Table I contains valuable crossover information but is not referenced in the main text where the 15-20 atom crossover is claimed; please add a pointer so the parameter dependence of the crossover is transparent to readers.","section":"Appendix B, Table I"}],"recommendation":"major_revision","confidential_remarks":"The core construction is sound and the algorithmic contributions are real, but the central quantitative claims are heavily dependent on empirical fits and on the unproven constancy of n_kappa. I recommend major revision rather than rejection because the issues are fixable within the manuscript's scope: a clear statement of what is proven versus fitted, error bars and sensitivity analysis for the exponents and crossover, and a benchmark against the best molecular-orbital baseline would substantially strengthen the paper. Please also check the consistency of the reported baseline scaling with the low-rank molecular-orbital results cited in the text."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe DG blocking construction is the real thing: blockwise SVD compression of a primitive diagonal basis against an active space is a new idea, and the block-diagonal sparsity of the two-electron tensor follows cleanly from the definitions. The specialized swap networks are a legitimate algorithmic contribution. The paper also does something rare: it tests the construction in two different primitive regimes (plane-wave dual and Gausslet) and reports DMRG speedups that look credible.\n\nWhere the paper is solid, it is solid. The count L = O(N_b^2 n_kappa^4) is derived correctly, and the numerical data support the qualitative claim that DG interpolates between diagonal and compact bases. I agree with the reader that there is no red flag for circularity or data fitting in the citations.\n\nThe soft spots are exactly where the report points. The headline O(N^2.6) cost claim depends on n_kappa staying bounded as system size grows at fixed accuracy. The paper itself says this is an expectation, not a theorem, and the supporting numerics are hydrogen chains up to H30 at fixed SVD tolerance, not fixed physical accuracy. The fitted exponents drift with tau: L exponents go 2.03, 2.18, 2.34, and lambda exponents go 1.42, 1.47, 1.58, so the cost exponent is really a range around 2.6. That does not sink the method. The crossover at 15–20 atoms is based on actual integral and lambda tallies, so it is more robust than the asymptotic fits. But the headline scaling should be read as suggestive, not demonstrated.\n\nThe paper is honest about its assumptions in the text, and the limitations are stated rather than hidden. It deserves a serious referee. In peer review I would ask for (1) code and data release, and (2) an analysis tying the SVD cutoff to a physical accuracy measure, plus at least one test beyond hydrogen chains to probe the n_kappa assumption. My recommendation is to send it out, expecting major revision or conditional acceptance.\n\nBring it to reading group. It will generate useful discussion about basis representations and what counts as evidence for asymptotic claims.","headline":"A genuinely new basis construction with a sound core, but the headline scaling claims rest on an unproven locality assumption and fits that shift with tolerance.","tokens_in":27571,"tokens_out":2383,"would_cite":true,"duration_ms":23812,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Block-diagonal basis cuts quantum chemistry cost to O(N^2.6)","keywords":["discontinuous Galerkin","block-diagonal two-electron integrals","quantum simulation","basis set discretization","swap networks","Gausslets","plane wave dual basis","hydrogen chains"],"falsifier":"Run the same DG blocking on a delocalized or strongly correlated model, such as a uniform electron gas or a hydrogen chain at large bond stretching, with a fixed SVD tolerance, and check whether the average number of DG functions per block converges as the number of atoms grows; if it grows without bound for fixed accuracy, the core scaling claim fails.","tokens_in":2171,"feed_emoji":"🧪","tokens_out":4900,"duration_ms":104153,"temperature":0.7,"pith_summary":"This paper constructs a basis that interpolates between compact but dense molecular-orbital bases and diffuse but diagonal grid bases. It partitions a diagonal primitive basis (plane-wave dual functions or Gausslets) into spatial blocks and compresses each block's projection onto a target active space with a singular value decomposition. The resulting discontinuous Galerkin basis has a block-diagonal two-electron tensor with $O(N_b^2 n_\\kappa^4)$ non-zero terms, so when the per-block function count $n_\\kappa$ stays bounded the integral count is $O(N_d^2)$. For hydrogen chains the paper finds that fault-tolerant quantum evolution cost drops from roughly $O(N^{4.5})$ in a Gaussian active space to $O(N^{2.6})$ in this basis, with a crossover around 15 to 20 atoms. DMRG calculations in the basis reach near-complete-basis-set accuracy with 1 to 2 orders of magnitude speedup.","feed_headline":"Block-diagonal basis cuts quantum chemistry cost to O(N^2.6)","feed_subtitle":"Compressing each spatial block keeps integrals sparse; the crossover beats Gaussian bases before 20 atoms.","key_machinery":"The central object is the blockwise singular value decomposition of the primitive-to-active-space coefficient matrix $\\Phi$. Partitioning the primitive index set into blocks $\\kappa$, writing the block restriction as $\\Phi_\\kappa \\approx U_\\kappa S_\\kappa V_\\kappa^\\dagger$, and keeping the leading $n_\\kappa$ left singular vectors defines DG functions $\\varphi_{\\kappa,j}(r)=\\sum_{\\mu\\in\\kappa}\\chi_\\mu(r)(U_\\kappa)_{\\mu,j}$. Because each block rotates only within itself, the two-electron tensor inherits the diagonal form of the primitive basis between different blocks, producing the block-diagonal sparsity pattern. This pattern is what the swap-network Trotter circuits and the LCU cost model exploit.","core_discovery":"The central claim is that an arbitrary active-space basis can be re-expressed in a basis of spatially blocked functions so that the two-electron integral tensor is exactly block diagonal: $v_{\\kappa,i;\\kappa',i';\\lambda,j;\\lambda',j'} = v^{(d)}_{\\kappa,\\kappa';i,i',j,j'}\\delta_{\\kappa\\lambda}\\delta_{\\kappa'\\lambda'}$. Each DG function is built from primitive functions in one block by retaining the leading $n_\\kappa$ left singular vectors of the primitive-to-active-space matrix restricted to that block. The block-diagonal form turns the quartic integral count into $O(N_b^2 n_\\kappa^4)$, which becomes $O(N_d^2)$ when $n_\\kappa$ is bounded by a constant. On hydrogen chains the empirical fault-tolerant evolution cost improves from $O(N^{4.5})$ to $O(N^{2.6})$, with the crossover in non-zero integrals and the $\\lambda$ factor occurring before or around 15 to 20 atoms. The same locality and block structure preserves accuracy in coupled-cluster and DMRG calculations, and yields large classical speedups.","pith_inferences":["One testable extension is to iterate the construction self-consistently: use a trial density matrix to define the active space, build the DG basis, recompute the density, and repeat; the paper's hybrid weighting recipe suggests this tuning could be made systematic.","If the per-block function count remains bounded in higher dimensions, the block-diagonal sparsity should transfer to plate or bulk systems; numerical checks on two-dimensional hydrogen clusters would test whether the chain results generalize.","The block-local form should reduce the entanglement burden of any tensor-network method, not only DMRG, because inter-block correlation is carried only through block-interaction terms; tree tensor networks might inherit a similar speedup.","The LCU estimate counts all non-zero integrals, but the block structure may allow further compression through blockwise low-rank factors, which would lower constants beyond the quoted crossover; the paper notes but does not quantify this possibility."],"forward_implications":["For fault-tolerant LCU simulation of hydrogen chains, the DG basis changes the empirical cost from about $O(N^{4.5})$ in a Gaussian active space to $O(N^{2.6})$, with the crossover in non-zero integral count and in $\\lambda$ appearing before or around 15 to 20 atoms.","For Trotter-based simulation, the block-diagonal structure yields swap-network depth $O(N_b n_\\kappa^3)=O(N_d n_\\kappa^2)$ for the quartic terms, interpolating between the diagonal linear-depth case and the dense cubic-depth case.","For DMRG, DG bases built on Gausslets with a hybrid active space reach near-complete-basis-set accuracy with one to two orders of magnitude lower cost than the primitive Gausslet or Gaussian basis alone.","The construction has a one-parameter family: setting $n_\\kappa=1$ gives a strictly diagonal primitive basis, while a single block reproduces the dense active-space Hamiltonian, so the method interpolates between the two regimes.","The DG representation also lowers the $\\lambda$ factor compared with the Gaussian active space on the tested chains (roughly $\\lambda\\propto N^{1.5}$ versus $N^{2.5}$), which benefits both fault-tolerant cost and measurement counts in variational algorithms."],"supporting_citations":[{"why":"Supplies the Gausslet primitive basis with approximate delta-function and strict localization properties used for the correlated DG calculations.","marker":"[34]"},{"why":"Provides the multisliced Gausslet basis and complete-basis-set benchmark data used to validate accuracy of DMRG in the DG basis.","marker":"[35]"},{"why":"Introduces the plane-wave dual basis and diagonal-Hamiltonian quantum simulation methods that serve as a primitive basis and a cost baseline.","marker":"[36]"},{"why":"Supplies the qubitization LCU cost model and prepare-oracle analysis used to estimate fault-tolerant scaling in arbitrary basis sets.","marker":"[38]"},{"why":"Gives the low-rank factorization of the Coulomb operator used in the alternative Trotter implementation within DG blocks.","marker":"[39]"},{"why":"Develops the linear-depth fermionic swap network for diagonal Hamiltonians that underlies the block-diagonal swap-network construction.","marker":"[57]"},{"why":"Provides the generalized swap-network primitives and hypergraph acquaintance strategy used to implement block-diagonal Hamiltonians in depth $O(N_b n_\\kappa^3)$.","marker":"[59]"},{"why":"Supplies the QROM data-lookup technique that determines the T-gate cost of the LCU prepare oracle as a function of the number of unique Hamiltonian coefficients.","marker":"[67]"},{"why":"Establishes the $O(N_p^{1/3})$ first-quantization scaling for diagonal basis sets that motivates the search for a block-diagonal middle ground.","marker":"[40]"},{"why":"Introduces the adaptive local basis set in a discontinuous Galerkin framework that this work adapts to correlated electronic structure and quantum computing.","marker":"[43]"}],"fun_headline_variants":["Block-diagonal basis lowers quantum cost to O(N^2.6)","Sparse block-diagonal basis beats Gaussians by 20 atoms","Quantum chemistry cost drops to O(N^2.6) with block-diagonal basis","Block diagonalization reduces quantum cost from O(N^4.5) to O(N^2.6)"],"cache_read_input_tokens":29696,"weakest_assumption_plain":"The paper's scaling results assume that a fixed small number of basis functions per spatial block is enough to keep accuracy as the molecule grows; if that number must grow with system size for delocalized or strongly correlated systems, the claimed $O(N^2)$ integral scaling and the $O(N^{2.6})$ quantum cost would both degrade.","fun_headline_variants_meta":{"raw":{"variants":["Block-diagonal basis lowers quantum cost to O(N^2.6)","Sparse block-diagonal basis beats Gaussians by 20 atoms","Quantum chemistry cost drops to O(N^2.6) with block-diagonal basis","Block diagonalization reduces quantum cost from O(N^4.5) to O(N^2.6)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001059,"raw_usage":{"total_tokens":4502,"prompt_tokens":1060,"completion_tokens":3442,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":676,"completion_tokens_details":{"reasoning_tokens":3354}},"tokens_in":676,"tokens_out":3442,"duration_ms":23516,"temperature":1.0,"reasoning_tokens":3354,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:03:59.908385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same DG blocking on a delocalized or strongly correlated model, such as a uniform electron gas or a hydrogen chain at large bond stretching, with a fixed SVD tolerance, and check whether the average number of DG functions per block converges as the number of atoms grows; if it grows without bound for fixed accuracy, the core scaling claim fails.","supporting_citations":[{"cited_title":"Gaussian basis sets for use in correlated molecular calculations. iii. the atoms aluminum through argon,","cited_arxiv_id":null,"evidence_quote":"Supplies the Gausslet primitive basis with approximate delta-function and strict localization properties used for the correlated DG calculations."},{"cited_title":"Linear scaling electronic structure meth- ods,","cited_arxiv_id":null,"evidence_quote":"Provides the multisliced Gausslet basis and complete-basis-set benchmark data used to validate accuracy of DMRG in the DG basis."},{"cited_title":"Ahlrichs, H","cited_arxiv_id":null,"evidence_quote":"Introduces the plane-wave dual basis and diagonal-Hamiltonian quantum simulation methods that serve as a primitive basis and a cost baseline."},{"cited_title":"An efficient and near linear scaling pair natural orbital based local coupled cluster method,","cited_arxiv_id":null,"evidence_quote":"Supplies the qubitization LCU cost model and prepare-oracle analysis used to estimate fault-tolerant scaling in arbitrary basis sets."},{"cited_title":"Sliced basis density matrix renormalization group for electronic structure,","cited_arxiv_id":null,"evidence_quote":"Gives the low-rank factorization of the Coulomb operator used in the alternative Trotter implementation within DG blocks."},{"cited_title":"Adaptive local basis set for kohn–sham density func- tional theory in a discontinuous galerkin framework ii: Force, vibration, and molecular dynamics calculations,","cited_arxiv_id":null,"evidence_quote":"Develops the linear-depth fermionic swap network for diagonal Hamiltonians that underlies the block-diagonal swap-network construction."},{"cited_title":"Finite element methods in ab initio electronic structure calculations,","cited_arxiv_id":null,"evidence_quote":"Provides the generalized swap-network primitives and hypergraph acquaintance strategy used to implement block-diagonal Hamiltonians in depth $O(N_b n_\\kappa^3)$."},{"cited_title":"Quantum Algorithm Pro- viding Exponential Speed Increase for Finding Eigenval- ues and Eigenvectors,","cited_arxiv_id":null,"evidence_quote":"Supplies the QROM data-lookup technique that determines the T-gate cost of the LCU prepare oracle as a function of the number of unique Hamiltonian coefficients."},{"cited_title":"Hybrid grid/basis set discretizations of the schr¨ odinger equation,","cited_arxiv_id":null,"evidence_quote":"Establishes the $O(N_p^{1/3})$ first-quantization scaling for diagonal basis sets that motivates the search for a block-diagonal middle ground."},{"cited_title":"Encoding electronic spectra in quantum circuits with linear t com- plexity,","cited_arxiv_id":null,"evidence_quote":"Introduces the adaptive local basis set in a discontinuous Galerkin framework that this work adapts to correlated electronic structure and quantum computing."}],"review_version":1}