{"id":"3fc5b4b9-7360-4323-a593-c96ded1f4721","arxiv_id":"1908.02456","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A discrete Dirichlet-to-Neumann map, derived from the semi-discrete Schrödinger equation and approximated by rational functions, yields stable absorbing boundary conditions for 3D quantum dynamics simulations.","lead":"This paper derives absorbing boundary conditions for three-dimensional Schrödinger-type equations from the finite-difference scheme itself, producing a discrete Dirichlet-to-Neumann map that is then approximated by rational functions. The result gives a systematic, less hand-tuned way to let outgoing quantum waves leave a simulation box.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem III.4's first-order stability proof silently assumes H_{II,Γ} has full column rank so Q=(H_{Γ,II}H_{II,Γ})^{-1} exists; the paper never states or verifies this rank condition.","rationale":"The most load-bearing concern is the unstated rank/invertibility assumption in Theorem III.4. The proof's Lyapunov argument depends essentially on Q=(H_{Γ,II}H_{II,Γ})^{-1}: without this matrix, the auxiliary term f^*Qf in the Lyapunov functional is not defined, the identity iI+QA=0 cannot be enforced, and the projection identity H_{II,Γ}(H_{Γ,II}H_{II,Γ})^{-1}H_{Γ,II} used to show negative semidefiniteness has no meaning. The paper neither states full column rank as a hypothesis nor verifies it for the geometries tested, so the theorem as written is incomplete, even though the numerical tests are consistent with the condition holding. This is a correctness gap in a central claim, not merely a missing experiment. The TDHF exterior-potential approximation is a second limitation, but it is acknowledged and is an application-level simplification rather than a flaw in the derivation. Because the rank condition likely holds for all tested grids and can be added as an explicit hypothesis or verified computationally, the concern does not overturn the paper's core derivation; it reinforces the conditional verdict. I therefore recommend keeping the reader's CONDITIONAL verdict unchanged.","tokens_in":21496,"tokens_out":11396,"duration_ms":116045,"concrete_test":"Reconstruct H_{II,Γ} for the 3D 7-point stencil on the 31^3 box used in Sec. IV.B and compute the smallest singular value of H_{Γ,II}H_{II,Γ}. If it is zero, Q is undefined and Theorem III.4 does not cover first-order stability on that grid. Repeat the computation on a non-rectangular domain (e.g., a sphere-like set of grid points) to probe the 'arbitrary geometry' claim; a zero or near-zero singular value would show the missing rank hypothesis is practically relevant.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central stability guarantee for the recommended first-order ABC (Theorem III.4, Sec. III.B) constructs Q = (H_{Γ,II}H_{II,Γ})^{-1} and asserts 'Q is symmetric positive definite.' This requires H_{II,Γ} (equivalently H_{Γ,II}H_{II,Γ}) to be invertible, i.e., H_{II,Γ} must have full column rank n_Γ. No hypothesis in the theorem or elsewhere states or verifies this condition. For the simple 1D 5-point stencil the matrix is block-triangular with nonzero diagonal, so the rank condition holds; for the box grids in Sec. IV it likely holds. But the paper claims the stability result and the method apply to arbitrary geometry, and the definition of Γ alone does not rule out rank-deficient exterior couplings (e.g., boundary sites whose exterior neighbors are linearly dependent). If the rank condition fails, K(s_1) may not be invertible, Q is undefined, the Lyapunov functional W = φ^*φ + f^*Qf is not positive definite, and the negative-semidefiniteness of QB+B^*Q cannot be established. The proof therefore secures first-order stability only under an unstated and unverified rank assumption. Separately, the TDHF application assumes ρ=0 in Ω_II so the long-range Yukawa/Coulomb potentials are dropped from the exterior; the paper labels this an expectation, not a proof (end of Sec. IV.C). The rank gap is more load-bearing because it directly undermines the theorem cited as the stability guarantee.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives absorbing boundary conditions for time-dependent Schrödinger-type equations in three dimensions, starting from a semi-discrete finite-difference model on an unbounded domain. It partitions the unknowns into interior and exterior blocks, defines a boundary layer Γ, and derives the exact discrete Dirichlet-to-Neumann map K(s) = -H_{Γ,II}[H_{II,II}-isI]^{-1}H_{II,Γ} in the Laplace domain (Eq. (16)). This map is then approximated by rational functions R_{m,m}(s), which yield local-in-time ODEs for the auxiliary boundary variables. Stability proofs are given for the zeroth-order (Theorem III.2) and first-order (Theorem III.4) approximations, while higher-order stability is treated numerically. The methods are tested on 1D and 3D linear Schrödinger benchmarks and on a 3D time-dependent Hartree-Fock 16O+16O collision with Skyrme interactions, comparing density, particle number, and energy against large-domain reference solutions.","tokens_in":21847,"tokens_out":8622,"duration_ms":92517,"significance":"If the stability and accuracy claims hold, the paper makes a useful contribution: it constructs boundary conditions directly from the discrete Hamiltonian used in the interior, so no further spatial discretization of the ABC is needed; the DtN map in Eq. (16) is exact model reduction for the semi-discrete problem, not a heuristic; and the rational approximation converts the nonlocal time convolution into a small ODE system. The numerical demonstration on a realistic TDHF model with Skyrme interactions is a strength, as are the explicit comparisons of different interpolation points and the connection of the first-order ABC to a parameter-free complex absorbing potential. However, the rigor of the stability claims is not yet at the level stated in the abstract and Section V: the first-order theorem requires an unstated rank condition, the zeroth-order proof contains an internal inconsistency, and the scheme implemented in the numerical experiments is not exactly the one covered by the first-order stability theorem. These issues are fixable within the scope of the paper, but they must be addressed before the central stability assertions can be accepted as proved.","major_comments":[{"comment":"The proof constructs Q=(H_{Γ,II}H_{II,Γ})^{-1} and asserts that Q is symmetric positive definite. This requires H_{Γ,II}H_{II,Γ} to be invertible, equivalently H_{II,Γ} must have full column rank n_Γ. The theorem neither states nor verifies this condition. If the boundary coupling block is rank-deficient, then A=-iH_{Γ,II}H_{II,Γ} is not invertible, Q is not positive definite, and the projection H_{II,Γ}(H_{Γ,II}H_{II,Γ})^{-1}H_{Γ,II} used in Eq. (34) is not an orthogonal projection, so the Lyapunov argument breaks down. The statement in Section V that the stability results do not depend on the geometry is therefore not justified. Please add the rank condition as an explicit hypothesis, prove it for the 7-point and 9-point stencils and box geometries used in Section IV, or restrict the theorem accordingly.","section":"Section III.B, Theorem III.4, Eqs. (32)-(34)"},{"comment":"The zeroth-order stability proof is internally inconsistent. It begins with 'We restrict M to be a Hermitian matrix,' but then concludes stability from the condition that 'M has a negative definite imaginary part,' and identifies M with K(s0). A Hermitian matrix has zero imaginary part, while K(s0) in Eq. (16) is non-Hermitian. The intended argument can be repaired by showing directly that K(s0) has negative-semidefinite imaginary part from the identity displayed in the proof, but as written the theorem conflates a Hermitian M with the non-Hermitian DtN map K(s0). Please remove the Hermitian restriction and state the sign condition on Im M explicitly.","section":"Section III.A, Theorem III.2"},{"comment":"The first-order stability theorem uses interpolation conditions at s=s1 (finite, positive) and at infinity via lim_{λ→0} (d/dλ)K. The numerical first-order ABCs, however, are implemented with two finite interpolation points: s=10 and 20 in Section IV.A, s=1 and 2 in Section IV.B, and pairs (s1,s2) in Section IV.C. The sentence after Theorem III.4 acknowledges only numerically that stability persists for finite s2. Consequently, the theorem does not cover the scheme that is actually tested and recommended. Please either prove stability for two finite interpolation points under explicit conditions, or clearly separate the proved variant from the numerically validated variant.","section":"Section III.B compared with Sections IV.A-IV.C"},{"comment":"The DtN derivation in Section II assumes that the exterior Hamiltonian is constant (V=0 or a constant) in Ω_II. In the TDHF application, the one-body Hamiltonian (49) contains density-dependent, Yukawa, and Coulomb terms that are nonlocal and long-range. The paper sets ρ=0 in Ω_II at t=0 and expects ρ≈0 in the exterior throughout, but this expectation is not proved or monitored. If particles or density reach the boundary, the ABC omits these potential terms and the error could grow. The manuscript explicitly labels this as an expectation rather than a proof. Please state this as an explicit limitation and provide a diagnostic, such as the maximum exterior density over the simulation time, or quantify the neglected exterior potentials.","section":"Section IV.C, Eqs. (49)-(53) and final remark"}],"minor_comments":[{"comment":"The displayed dimensions of E are inconsistent with the explicit form E=[I_{n_Γ} 0]; the matrix should map from R^{n_I} to R^{n_Γ}, not from R^{n_{II}} to R^{n_Γ}.","section":"Section II, Eq. (7)"},{"comment":"The text refers to 'the DtN map K(t) is a 14166×14166 dense matrix'; since K is defined in the Laplace domain, this should read K(s), not K(t).","section":"Section IV.B, first paragraph"},{"comment":"The notation 'K(s): R → Ω_Γ × Ω_Γ' is not correct: K(s) is a matrix-valued function on C with values in C^{n_Γ×n_Γ}, and Ω_Γ is a set of grid points. Please adjust the notation.","section":"Section II, after Eq. (16)"},{"comment":"The phrase 'where s1 is a any positive real number' contains a typo ('a any'); it should be 'any positive real number.'","section":"Section III.B, Theorem III.4 statement"},{"comment":"The assumption is written as 'ψ_j = 0 for i=1,...,A and ρ=0 in Ω_II', but the index in the first clause should be j, not i.","section":"Section IV.C, Eq. (53)"},{"comment":"The symbol c_k is used both for the finite-difference coefficients and for the grid offsets in the exponential; please use a separate symbol for one of them to avoid confusion.","section":"Appendix D, Eq. (D4)"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my read of arXiv:1908.02456.\n\nThe genuinely new piece is the discrete DtN map for the semi-discrete Schrödinger equation. By starting from the finite-difference Hamiltonian before deriving the boundary condition, Wu and Li get an ABC that matches high-order stencils without extra spatial discretization. The Laplace-domain rational approximation with ODE realization is the practical move, and the Lyapunov stability proofs for orders zero and one are a real contribution. The 1D and 3D tests are reasonable, and the TDHF example goes beyond the usual toy problems. The citation pattern looks honest; the prior discrete-DtN work on acoustic waves and their own molecular-dynamics ABC are acknowledged.\n\nThe main soft spot is Theorem III.4. The proof chooses Q=(H_{Γ,II}H_{II,Γ})^{-1} and asserts it is symmetric positive definite. That requires H_{II,Γ} to have full column rank, i.e., the boundary-to-exterior coupling block must be injective. The paper never states or checks this. For the structured grids used in the numerical section it almost certainly holds, so I do not see a counterexample to the method; but as written the stability theorem covers only cases where that rank condition is true. If the authors claim arbitrary geometry, they need to state the assumption and either prove it for reasonable grids or verify it in each example. This is patchable, but it is the load-bearing gap.\n\nThe other weaknesses are in proportion smaller. The 'arbitrary geometry' claim is supported by no non-box test; the DtN evaluation via boundary elements is described but not validated on a curved or irregular domain. The TDHF application drops the long-range Coulomb/Yukawa terms in the exterior, justified only as 'we expect ρ≈0'; that is an assumption, not a proof, and it should be acknowledged more prominently. Finally, the numerical verification has no convergence study: no h-refinement, no reflection-coefficient measurements, and the 3D 'exact' solution is the continuum Gaussian, not the solution of the discrete model, so the comparison is qualitative. Those are the things a referee should push on.\n\nAll in all, this paper deserves a serious referee. The discrete-DtN derivation and stability results are solid enough to be worth engaging with. I would send it out, with the expectation that the rank condition, a non-box geometry test, and some convergence data get added. For my own work, I would cite it if I were writing about absorbing boundary conditions for Schrödinger-type equations.","headline":"A genuinely useful discrete-DtN absorbing boundary condition with real stability theorems, but the first-order proof hides a rank assumption and the arbitrary-geometry claim outruns the experiments.","tokens_in":22335,"tokens_out":5687,"would_cite":true,"duration_ms":64567,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M06","65M12","65M80","35Q41"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper derives absorbing boundary conditions for 3D Schrödinger-type equations from a discrete Dirichlet-to-Neumann map, proves stability for the zeroth and first orders, and gets second-order accuracy close to exact solutions.","keywords":["absorbing boundary conditions","Dirichlet-to-Neumann map","time-dependent Schrödinger equation","finite-difference discretization","rational approximation","stability analysis","time-dependent Hartree-Fock","three-dimensional simulation"],"falsifier":"Form the boundary-coupling product $H_{\\Gamma,II}H_{II,\\Gamma}$ for a cubic grid with the 7-point Laplacian stencil and test whether it is nonsingular; if it is singular, the Lyapunov matrix for the first-order ABC does not exist, and a numerical run of the 16O+16O TDHF model with nucleons reaching the boundary would show whether energy or nucleon number grows instead of decaying.","tokens_in":21300,"feed_emoji":"⚛️","tokens_out":7922,"duration_ms":80052,"temperature":0.7,"pith_summary":"The paper's goal is to make absorbing boundary conditions practical for three-dimensional time-dependent Schrödinger-type equations by deriving them from the finite-difference discretization itself, rather than from the continuous PDE. It shows that the exterior region's influence on the computational domain can be written exactly as a discrete Dirichlet-to-Neumann map, a matrix-valued function that maps boundary values to the flux entering the domain. The central claim is that rational approximations of this map yield boundary conditions that are provably stable at zeroth and first order, and numerically accurate at second order across 1D, 3D, and time-dependent Hartree-Fock tests. If correct, this gives quantum simulations a way to cut the computational domain without introducing a hand-tuned absorbing layer.","feed_headline":"Stable absorbing boundaries for 3D Schrödinger equations","feed_subtitle":"Rational approximations of an exact boundary map let quantum waves leave the simulation box with little reflection.","key_machinery":"The central object is the matrix-valued discrete DtN map in Eq. (16), $K(s) = -H_{\\Gamma,II}[H_{II,II}-isI]^{-1}H_{II,\\Gamma}$, where $H_{\\Gamma,II}$ and $H_{II,\\Gamma}$ are the boundary-coupling blocks of the finite-difference Hamiltonian and $H_{II,II}$ is the exterior block. The paper replaces this map with rational interpolants $R_{m,m}(s) = (s^mI - s^{m-1}B_0 - \\cdots - B_{m-1})^{-1}(s^{m-1}A_0 + \\cdots + A_{m-1})$; in the time domain the interpolation coefficients become coefficients of an ODE system, so the convolution becomes cheap to evolve. Evaluating $K(s)$ at interpolation points is done with discrete Green's functions and selected inversion, reducing the problem to small boundary blocks even for arbitrary geometry. The stability arguments use a Lyapunov functional, with the first-order proof choosing $Q = (H_{\\Gamma,II}H_{II,\\Gamma})^{-1}$ as the weighting matrix.","core_discovery":"The paper establishes that for the semi-discrete Schrödinger equation, the exact effect of the exterior on the interior is the Laplace-domain operator $K(s) = -H_{\\Gamma,II}[H_{II,II}-isI]^{-1}H_{II,\\Gamma}$, called the discrete Dirichlet-to-Neumann map. It approximates $K(s)$ by rational matrix functions, whose inverse Laplace transform turns the nonlocal time convolution into linear ODEs for a boundary flux variable. Theorems III.2 and III.4 prove stability of the zeroth- and first-order approximations under stated conditions on the interpolation points, and numerical tests in section IV show that the second-order ABC nearly matches exact solutions in 1D and 3D, including at corners, and that the ABC releases nucleons and energy in the 16O+16O time-dependent Hartree-Fock model in good agreement with the large-domain reference.","pith_inferences":["A direct extension the paper does not pursue is applying the same rational-interpolation boundary condition to time-dependent density-functional theory, where the exterior Hamiltonian is again the free Laplacian, so the DtN map would be unchanged.","The TDHF sweep suggests the interpolation scale near $s=1$ is optimal for a 30-fm box; a testable conjecture is that the optimal scale is set by the inverse box size, letting users choose points a priori.","The missing rank verification in the stability proof is likely to hold for standard 7-point and 9-point stencils on Cartesian grids, but it should be checked for coarse, anisotropic, or unstructured discretizations before applying the first-order ABC there."],"forward_implications":["The zeroth-order ABC is equivalent to a complex absorbing potential derived from the exact DtN map, so it needs no empirical choice of absorbing strength.","The first-order ABC replaces the nonlocal time convolution with a small system of ordinary differential equations for boundary auxiliary variables.","Because the DtN map is evaluated through discrete Green's functions, the method works for arbitrary domain geometry, including corners and edges.","Numerical tests show that increasing the approximation order reduces reflection, with the second-order ABC nearly matching the exact solution in 1D and 3D.","In the 16O+16O time-dependent Hartree-Fock test, the ABC lets nucleons and energy leave the box, with the best cases closely tracking the exact large-domain energy and density."],"supporting_citations":[{"why":"Supplies the method of starting from a spatially discrete model to derive a discrete DtN map for unbounded domains.","marker":"[53]"},{"why":"Provides the discrete boundary-element technique used to evaluate the DtN map through Green's functions.","marker":"[49]"},{"why":"Supplies the authors' previous stable ABC construction and discrete Green's function evaluation used in the DtN map implementation.","marker":"[66]"},{"why":"Describes the selected-inversion algorithm that makes computing the needed entries of the inverse feasible.","marker":"[58]"},{"why":"Provides the complex absorbing potential used as a baseline comparison in the 1D test.","marker":"[14]"},{"why":"Supplies the 3D Schrödinger benchmark problem used to test the ABCs.","marker":"[40]"},{"why":"Analyzes ABC stability for molecular dynamics with Lyapunov functionals, the approach adapted for the stability theorems.","marker":"[62]"},{"why":"Provides the 16O+16O time-dependent Hartree-Fock model and data used in the application test.","marker":"[63]"}],"fun_headline_variants":["Exact DtN map yields stable 3D Schrödinger ABCs","Rational ABCs for 3D Schrödinger equations","Stable absorbing boundaries for 3D quantum simulations","Leak-free boundaries for 3D Schrödinger-type equations","Rational approximations shrink Schrödinger simulation costs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The first-order stability proof relies on the product of the boundary-coupling matrices being invertible and positive definite; the paper states this but never verifies the rank condition, so a singular coupling would leave the proof without a Lyapunov function.","fun_headline_variants_meta":{"raw":{"variants":["Exact DtN map yields stable 3D Schrödinger ABCs","Rational ABCs for 3D Schrödinger equations","Stable absorbing boundaries for 3D quantum simulations","Leak-free boundaries for 3D Schrödinger-type equations","Rational approximations shrink Schrödinger simulation costs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000317,"raw_usage":{"total_tokens":1772,"prompt_tokens":901,"completion_tokens":871,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":791}},"tokens_in":517,"tokens_out":871,"duration_ms":8682,"temperature":1.0,"reasoning_tokens":791,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:44:58.878632+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Form the boundary-coupling product $H_{\\Gamma,II}H_{II,\\Gamma}$ for a cubic grid with the 7-point Laplacian stencil and test whether it is nonsingular; if it is singular, the Lyapunov matrix for the first-order ABC does not exist, and a numerical run of the 16O+16O TDHF model with nucleons reaching the boundary would show whether energy or nucleon number grows instead of decaying.","supporting_citations":[{"cited_title":"Implementation of transparent boundaries for numerical solution of the Schr¨ odinger equation","cited_arxiv_id":null,"evidence_quote":"Supplies the method of starting from a spatially discrete model to derive a discrete DtN map for unbounded domains."},{"cited_title":"Artificial boundary conditions for one-dimensional cubic nonlinear Schr¨ odinger equations","cited_arxiv_id":null,"evidence_quote":"Provides the discrete boundary-element technique used to evaluate the DtN map through Green's functions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the authors' previous stable ABC construction and discrete Green's function evaluation used in the DtN map implementation."},{"cited_title":"Quarteroni and A","cited_arxiv_id":null,"evidence_quote":"Describes the selected-inversion algorithm that makes computing the needed entries of the inverse feasible."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the complex absorbing potential used as a baseline comparison in the 1D test."},{"cited_title":"Efficient represent ation of nonreflecting boundary con- ditions for the time-dependent Schr¨ odinger equation in tw o dimensions","cited_arxiv_id":null,"evidence_quote":"Supplies the 3D Schrödinger benchmark problem used to test the ABCs."},{"cited_title":"Pad´ e approximants, volume 59","cited_arxiv_id":null,"evidence_quote":"Analyzes ABC stability for molecular dynamics with Lyapunov functionals, the approach adapted for the stability theorems."},{"cited_title":"On the stability of boundary conditions for molecular dynamics","cited_arxiv_id":null,"evidence_quote":"Provides the 16O+16O time-dependent Hartree-Fock model and data used in the application test."}],"review_version":1}