{"id":"703bfcee-a393-4443-91bb-d7c24d629581","arxiv_id":"1909.01578","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Using Lebedev and Popov spherical grids in a discrete variable representation makes hydrogen-in-crossed-fields energies converge faster than with a product of Gaussians, and Popov is fastest.","lead":"A quantum-computation paper tests a faster angular grid for problems where the angular motion cannot be separated into independent directions. It finds that grids called Lebedev and Popov make the calculation converge much faster than the old grid, with Popov fastest.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eqs. (10)–(15) leave the high-l (l>lmax) quadrature error in L^2 unquantified; the claimed acceleration could saturate or misrank schemes if that error does not vanish with N.","rationale":"The paper's numerical demonstration on the hydrogen-in-crossed-fields problem is real evidence: the energies converge to the values of Melezhik [2], and the Popov/Lebedev grids reach a given accuracy with fewer basis functions than the product Gaussian grid. My concern is about the theoretical guarantee behind the method, not about the specific numbers. In a standard DVR the quadrature is supposed to be exact for products of basis functions; here the cubature is exact only for l ≤ lmax while the basis contains l ≤ ltop with ltop > lmax. Diagonalizing Eq. (10) makes (12)–(13) true in the discrete inner product, so the reader's phrasing of the assumption as 'orthogonal and complete on the grid' is satisfied by construction. The genuine risk is that Eq. (15) is not the exact Galerkin representation of L^2 and the quadrature error from the uncontrolled tail may not vanish as N grows. The ratio ltop/lmax ≈ 1.15 (from Table 1 asymptotics) means the tail is not a small fraction of the basis, so this is not a priori negligible. The proposed test directly compares the cubature L^2 with an exact reference and checks whether using the exact matrix changes the computed energies; this settles whether the concern actually lands. The reader's CONDITIONAL verdict remains appropriate, with the condition sharpened to require an error/conditioning analysis of the L^2 construction.","tokens_in":9530,"tokens_out":17664,"duration_ms":175839,"concrete_test":"Using the largest grids from Table 1 (Lebedev N=110, Popov N=104), compute the L^2 matrix of Eq. (15) twice: once with the cubature weights and once with a high-order product Gauss-Legendre reference quadrature converged to machine precision (e.g., Nθ=Nφ=400). Report max relative difference ‖L^2_cub − L^2_ref‖/‖L^2_ref‖. Then recompute the β=2.0, γ=1.0 ground-state energy replacing Eq. (15) by L^2_ref; if the energy shifts by more than the plotting tolerance in Fig. 1, the reported convergence is limited by cubature error in the high-l tail. Also report the smallest nonzero eigenvalue of the Nbar×Nbar overlap matrix (10) relative to the largest; if it is below ~1e-8, the corresponding basis function is numerically ill-defined.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The construction in Eqs. (10)–(15) diagonalizes the cubature-approximated overlap matrix and uses the resulting discrete-orthonormal functions as a DVR basis. Orthogonality and completeness on the grid, Eqs. (12)–(13), follow from the diagonalization; the unresolved issue is how faithfully that basis represents the continuous angular dynamics. The Lebedev/Popov cubatures are exact only for products of spherical harmonics with l ≤ lmax, whereas the basis includes l up to ltop > lmax (Table 1; e.g., N=110, lmax=8, ltop=12). Hence Eq. (15) for L^2 is a quadrature approximation whose error involves harmonics up to 2ltop. The paper neither bounds this quadrature error nor reports the conditioning of the overlap matrix. Because ltop/lmax tends to ≈1.15 for large N, the high-l tail remains a nonvanishing fraction of the basis; the paper does not show that its contribution to the quadrature error decays with N. If that error is large, the convergence curves in Fig. 1 could saturate or the ranking of the three schemes could change at stricter tolerances. This is the most load-bearing gap because both 'substantially accelerates' and 'fastest with Popov' presuppose that the constructed basis is a faithful angular discretization.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a nondirect product discrete variable representation (npDVR) for the angular part of the Schrödinger equation when the angular variables are nonseparable. The new basis is constructed by diagonalizing the overlap matrix of spherical harmonics computed on Lebedev or Popov cubature grids for the unit sphere, rather than on the direct product of one-dimensional Gaussian grids used previously. The authors test the method by computing bound states of a hydrogen atom in mutually tilted electric and magnetic fields, comparing the Gaussian-product, Lebedev, and Popov schemes. They report that the Popov-based npDVR converges fastest, followed by the Lebedev-based one, and they attribute this ordering to the cubature efficiency coefficient η.","tokens_in":9772,"tokens_out":6846,"duration_ms":71004,"significance":"If the reported acceleration is robust, the method is a genuinely useful tool for quantum dynamics with nonseparable angular degrees of freedom: it removes the polar clustering of the old product grid and can reach accurate energies with substantially fewer angular points. The paper has real strengths: the basis construction is described in enough detail to be reproduced, the hydrogen-in-crossed-fields test is a standard benchmark with an independent earlier value from Ref. [2], and the tables give the numerical data in full rather than as anonymous convergence plots. The main limitation is that the central correlation claim is supported by a single example and by heuristic reasoning, while the underlying quadrature-exactness issue for the high-l part of the basis is not analyzed.","major_comments":[{"comment":"The overlap matrix in Eq. (10) is evaluated with cubature weights that are exact only for products of spherical harmonics with angular momenta up to lmax, whereas the basis includes harmonics up to ltop > lmax (Table 1 gives, e.g., N=110 with lmax=8 and ltop=12, and N=104 with lmax=8 and ltop=11). The quadrature error in the Gram matrix and in the L2 matrix (15) is never quantified, and the condition number of the overlap matrix is not reported. Since ltop/lmax is not shown to tend to unity as N grows, the high-l component of the basis does not automatically become negligible; consequently it is not established that the quadrature error decays with N. This gap is load-bearing because a poor discrete representation of the high-l tail could cause the convergence curves in Fig. 1 to saturate or could change the relative ranking of the three schemes at stricter tolerances. The revision should either bound the quadrature error, report the spectrum and conditioning of the overlap matrix, or demonstrate numerically that the error decays with N.","section":"Sec. 2, Eqs. (10)-(15)"},{"comment":"The claimed direct correlation between η and convergence is not fully supported by the tabulated data because the convergence is non-monotonic in N. In Table 2 the Gaussian scheme gives -0.1307377 at N=49 but -0.1307454 at N=50 and N=70, and in Table 3 it gives -0.1304448 at N=30 and N=36 but -0.1302756 at N=40, followed by -0.1302755 at N=49 and -0.1302756 at N=50 and N=70. These jumps mean that comparisons at a fixed N can depend on the chosen tolerance and on which side of a jump the comparison is made. The paper should define a convergence criterion (e.g., deviation from a converged reference energy) and show that the ordering 'Gaussian slower than Lebedev slower than Popov' holds under that criterion, ideally for more than one state and more than one field configuration.","section":"Sec. 3, Tables 2 and 3"}],"minor_comments":[{"comment":"The thin horizontal lines labeled 'most accurate value' are apparently obtained from the large-N limit of the same schemes, but the caption does not state the reference value used or the tolerance for convergence; please make this explicit.","section":"Fig. 1"},{"comment":"The arrangement of the Gaussian columns for Nφ=3,5,7 is confusing, and the N=49/N=50 entries should be explained in terms of the underlying Nθ, Nφ grid parameters, especially because those two rows show the largest non-monotonic jumps.","section":"Tables 2 and 3"},{"comment":"The statement that the Lebedev or Popov cubature can exactly integrate 'the square of spherical harmonics up to a certain angular momentum (lmax)' is imprecise: the cubatures are exact for all products of pairs of spherical harmonics whose angular momenta do not exceed lmax, which is the property actually needed in Eq. (10).","section":"Sec. 2, text near Eq. (9)"},{"comment":"Several reference-list typos should be corrected: 'Phyladelphia' and 'Schience' in Ref. [45], 'editted' in Ref. [4], and 'Retreived' in Ref. [49].","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is squarely within the scope of physics.comp-ph and the basic idea is appealing. My main reservation is not about novelty but about the strength of the evidence for the central claim: the missing high-l quadrature-error analysis and the non-monotonic convergence in the tables make the 'fastest with Popov' statement less definitive than the abstract suggests. A revision that adds numerical diagnostics for the overlap matrix and a second test case, together with a more cautious wording of the correlation claim, would make the paper publishable. I do not see evidence of circularity or of data fabrication; the comparison with the independent value in Ref. [2] is a genuine strength."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it replaces the product-Gaussian angular grid in the nondirect product DVR with Lebedev and Popov cubatures, constructs the basis by diagonalizing the overlap matrix, and tests the scheme on hydrogen in crossed electric and magnetic fields. The new bit is the Popov cubature application and the direct comparison of all three grids. That is a legitimate extension of Haxton's earlier outline, and the numerical demonstration is clean: the Lebedev- and Popov-based schemes reach converged energies with far fewer points than the Gaussian scheme, and the weak-field values agree with Melezhik's published results. The orthogonalization procedure is clearly described and the efficiency coefficient eta is an independent property of each cubature, not a fitted parameter, so the circularity concern is minor.\n\nThe soft spots are real but proportionate. The most substantive one is the unquantified quadrature error for products of harmonics with l > lmax. The overlap matrix and the L^2 operator are evaluated with cubatures that are exact only up to n, while the basis includes harmonics up to ltop > n/2. The paper does not bound the error from these high-l products, and the stress-test note is right that the convergence curves could in principle saturate or even reorder at tighter tolerances. Still, the numerical results are consistent and agree with independent benchmarks, so this is a gap in analysis rather than a load-bearing flaw. I would want the authors to add a section estimating that error, at least numerically, and to comment on the conditioning of the overlap matrix.\n\nTwo smaller issues. First, the claimed direct correlation between eta and convergence is generalized from one physical problem. It is plausible and the heuristic argument is sensible, but it would be more convincing with a second example, e.g., a coupled-channel scattering problem or a time-dependent simulation. Second, the Gaussian scheme shows non-monotonic convergence in Figure 1, and the paper never addresses why; a sentence or two would help the comparison. Shipping the code would also have been nice, though by 2019 standards its absence is not disqualifying.\n\nWho gets value: anyone building angular DVRs for nonseparable problems, especially time-dependent two-active-electron simulations. The paper deserves a serious referee and, after the error analysis is added, could be accepted. I would take it if it crossed my desk, and I would probably cite it in my own work on angular discretizations.","headline":"A solid, incremental extension of npDVR to Lebedev and Popov cubatures with a convincing single-system benchmark, but the claimed correlation between cubature efficiency and convergence is not rigorously established and the quadrature error for the high-l tail is left unquantified.","tokens_in":10282,"tokens_out":1947,"would_cite":true,"duration_ms":23630,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A wave-function basis built on Popov or Lebedev unit-sphere cubatures converges substantially faster than the old Gaussian-product angular grid for Schrödinger equations with nonseparable angular variables.","keywords":["nondirect product DVR","Lebedev quadrature","Popov quadrature","unit-sphere cubature","angular discretization","hydrogen atom crossed fields","efficiency coefficient","Schrödinger equation"],"falsifier":"Take a fixed N-point Popov grid (e.g. N=12), build the overlap matrix (10) up to the stated ltop, and compare its eigenvalues with the ideal values 1 and 0; any nonzero eigenvalue beyond N, or significant deviation of the N kept eigenvalues from unity, would show the grid does not support the claimed completeness and should worsen convergence. The same comparison using a high-order reference quadrature for the exact scalar products would settle it.","tokens_in":9330,"feed_emoji":"⚛️","tokens_out":5107,"duration_ms":49521,"temperature":0.7,"pith_summary":"This paper develops a nondirect product discrete variable representation (npDVR) for quantum problems whose angular variables cannot be separated, replacing the old direct product of 1D Gaussian quadratures with Lebedev or Popov 2D cubatures on the unit sphere. It constructs orthogonalized basis functions on the cubature grid points and tests them on the hydrogen atom in arbitrarily oriented electric and magnetic fields, a genuinely nonseparable three-dimensional problem. The paper finds that both cubature-based schemes converge substantially faster than the Gaussian-product scheme at the same number of grid points, and that the Popov-based npDVR converges fastest, consistent with its higher efficiency coefficient. If correct, this gives quantum dynamics a cheaper angular discretization that also avoids the pole-clustering of product grids, making time-dependent few-electron simulations more feasible.","feed_headline":"Popov-grid DVR converges fastest for nonseparable angles","feed_subtitle":"Replacing product Gaussian grids with Lebedev or Popov cubatures cuts the grid size needed for hydrogen in crossed fields.","key_machinery":"The central object is the cubature-orthogonalized basis Φν(Ω)=∑μ SνμYμ(Ω), obtained by diagonalizing the overlap matrix built from spherical harmonics on the Lebedev or Popov grid points. The key identity is the orthogonality and completeness pair (12)-(13) on the grid, which makes the potential matrix diagonal and yields the L2 matrix formula (15). The paper ties the method's performance to the cubature efficiency coefficient η, the ratio of exactly integrated spherical harmonics to the number of grid degrees of freedom, and argues that larger η should give a better approximation for a fixed number of basis functions.","core_discovery":"The central claim is that a 2D npDVR built from invariant cubatures on the unit sphere outperforms the earlier npDVR built from a direct product of 1D Gaussian quadratures. The paper constructs the basis by extending the spherical-harmonic set until the overlap matrix (10), evaluated with the cubature weights, has exactly N nonzero eigenvalues; the normalized eigenvectors define functions Φν that are orthogonal and complete on the grid. With these functions the angular kinetic-energy operator L2 takes the explicit form (15) and the potential remains diagonal. Tested on the hydrogen atom in crossed magnetic and electric fields, the paper reports clearly faster convergence for both the Lebedev and Popov grids than for the Gaussian-product npDVR, with the Popov grid giving accurate results already at N=12. The paper concludes that convergence speed tracks the cubature efficiency coefficient η=(n+1)2/(3N).","pith_inferences":["The efficiency-coefficient heuristic could be used as a design rule for pruned angular grids in higher-dimensional problems, where direct-product grids become prohibitively large.","The overlap-matrix diagonalization used here can serve as a general diagnostic: before committing to dynamics on any candidate angular grid, one can check that the number of near-unit eigenvalues matches the grid size.","If the reported acceleration holds more broadly, similar gains should appear in time-dependent two-electron problems, a direction the paper only hints at in its conclusion.","The high-l tail of the extended harmonic set carries the quadrature error; testing the method on potentials with stronger high-l coupling would show how far the ltop extension can be pushed before the approximation degrades."],"forward_implications":["For the hydrogen spectrum in crossed fields, the Popov-based npDVR reaches accurate energies with far fewer angular grid points than the Gaussian-product scheme, with N=12 already giving good results.","The absence of pole clustering in the Lebedev and Popov grids removes a time-step restriction in time-dependent Schrödinger implementations.","The construction works with any invariant cubature on the unit sphere, so the efficiency-coefficient ranking provides a direct criterion for choosing the angular grid.","The method extends naturally to Schrödinger problems with nonseparable angular variables where partial-wave expansions become inefficient due to strong coupling.","The reported correlation between η and convergence gives a practical rule: for a fixed number of grid points, the cubature with the highest η should be preferred."],"supporting_citations":[{"why":"Supplies the Lebedev cubature point sets and weights used as one of the two new angular grids.","marker":"[19–23]"},{"why":"Supplies the Popov cubature rule whose grids give the fastest convergence reported.","marker":"[24]"},{"why":"Provides the overlap-matrix orthogonalization procedure the paper adapts to extend the harmonic basis to ltop.","marker":"[47]"},{"why":"Defines the efficiency coefficient η used throughout to rank quadratures and grids.","marker":"[43]"},{"why":"Gives the earlier npDVR calculation of the same hydrogen spectrum against which the new schemes are compared in Tables 2 and 3.","marker":"[2]"},{"why":"Establishes the original nondirect product DVR framework that this paper modifies.","marker":"[1–5]"}],"fun_headline_variants":["Popov-grid DVR: fastest convergence for nonseparable angles","Angular DVR gets speed boost from Popov cubature","Nonseparable angles? Popov grid converges faster","Lebedev and Popov grids speed up quantum DVR","New angular basis cuts grid size for crossed fields"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Lebedev or Popov cubature weights compute the spherical-harmonic overlap matrix accurately enough up to ltop so that the N kept eigenvectors really are orthogonal and complete on the grid; if the cubature errors for high-l products are not small, the whole npDVR basis is only approximately orthogonal.","fun_headline_variants_meta":{"raw":{"variants":["Popov-grid DVR: fastest convergence for nonseparable angles","Angular DVR gets speed boost from Popov cubature","Nonseparable angles? Popov grid converges faster","Lebedev and Popov grids speed up quantum DVR","New angular basis cuts grid size for crossed fields"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000239,"raw_usage":{"total_tokens":1486,"prompt_tokens":891,"completion_tokens":595,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":513}},"tokens_in":507,"tokens_out":595,"duration_ms":5858,"temperature":1.0,"reasoning_tokens":513,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:13:09.682284+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a fixed N-point Popov grid (e.g. N=12), build the overlap matrix (10) up to the stated ltop, and compare its eigenvalues with the ideal values 1 and 0; any nonzero eigenvalue beyond N, or significant deviation of the N kept eigenvalues from unity, would show the grid does not support the claimed completeness and should worsen convergence. The same comparison using a high-order reference quadrature for the exact scalar products would settle it.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Popov cubature rule whose grids give the fastest convergence reported."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the overlap-matrix orthogonalization procedure the paper adapts to extend the harmonic basis to ltop."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the efficiency coefficient η used throughout to rank quadratures and grids."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the earlier npDVR calculation of the same hydrogen spectrum against which the new schemes are compared in Tables 2 and 3."}],"review_version":1}