{"id":"af7bc6ca-c23b-46bd-b3d7-7817159883fa","arxiv_id":"1908.04234","paper_version":3,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A numerical method that excises the outer computational boundary along the ingoing light cone makes boundary conditions unnecessary, at the cost of limiting evolution to about one light-crossing time of the initial grid.","lead":"Numerical relativity simulations usually need careful boundary conditions at the edge of the grid. This paper shows that cutting away (excising) the outer boundary along the innermost light ray each timestep removes the need for boundary conditions, and demonstrates the idea with a spherically symmetric scalar field collapse code.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"CFL-obeying excision claim is sound; the CFL-violating variant rests on an unproven error-containment assumption that the paper itself flags.","rationale":"The reader's ACCEPT is correct. The central claim is the CFL-obeying excision method: moving the outer boundary inward at a speed at least |c_-| keeps every retained point's domain of dependence inside the old grid, so no ingoing characteristic crosses the boundary and no boundary condition is required. The code implements this and shows second-order convergence of the constraint residual E_theta theta. The weakest assumption identified by the reader concerns method II, which relaxes CFL by excising along the null ray. That assumption is indeed unproven and only tested in 1+1, but the paper states the limitation explicitly and the main numerical results do not depend on method II. Therefore the concern does not change the verdict. Agreement is partial because the reader treats this as the weakest assumption of the central claim, whereas it is best understood as a scoped extension rather than the load-bearing part of the main excision argument.","tokens_in":10616,"tokens_out":15752,"duration_ms":188747,"concrete_test":"Run method II in the same 1+1 code on a linear scalar wave equation with an analytic solution, such as an outgoing Gaussian pulse, and measure the interior L2 error at a fixed physical time for increasing resolution; if the boundary error layer does not shrink to zero width and amplitude with resolution, the containment assumption fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Sec. II A and Sec. III D introduce a second variant, method II, that excises directly along the ingoing null ray by removing one grid point every 1/(c_- lambda) time steps, deliberately violating the CFL condition at the boundary. The method is convergent only if the numerical errors generated by the missing ingoing data have characteristic speeds bounded by c_-, so that they remain in a layer near the moving boundary whose width tends to zero with resolution. This is asserted rather than proved, and the numerical support is a single 1+1 self-gravitating scalar in Painleve-Gullstrand coordinates; the paper explicitly cautions that other gauges and full 3+1 codes may behave differently. If the premise fails, method II would contaminate the interior and would not converge. This does not threaten the paper's main demonstration, which uses method I, the CFL-obeying excision every time step with boundary speed dx/dt at least |c_-|; for method I the domain-of-dependence argument is solid and the convergence tests support it. The soft spot is therefore real but scoped to the optional CFL-relaxation extension.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes an 'excision method' for evolving hyperbolic systems, in particular the Einstein equations, on a finite computational domain without imposing boundary conditions. At each time step the outer boundary is moved inward along a surface that is spacelike or tangent to the innermost characteristic, so that no characteristics enter the computational domain. The author contrasts this with the 'expansion method' of Bieri, Garfinkle, and Yau, discusses null-coordinate gauges in which the excision surface can be null, and describes a 1+1 code for a self-gravitating massless scalar in Painlevé-Gullstrand coordinates. Numerical examples cover both non-black-hole and black-hole-forming pulses. The CFL-obeying version (method I) converges at second order in the Einstein constraint residual (Figs. 6 and 8); a CFL-violating version (method II) that excises along the null ray is also tested and shown to have a boundary error layer whose width decreases with resolution, with an explicit caveat that this is only demonstrated in spherical symmetry.","tokens_in":10853,"tokens_out":8332,"duration_ms":90519,"significance":"If the central claim holds, the method offers a simple way to avoid well-posedness and constraint-preservation issues of initial-boundary value formulations for local collapse simulations (e.g., critical collapse or black-hole interiors), at the cost of a maximum evolution time of about one light-crossing time for explicit schemes. The domain-of-dependence argument for method I is standard and correct, and the Einstein-constraint residual is an independent check that gives clean second-order convergence in both non-black-hole and black-hole cases. The paper is honest about the heuristic status of the CFL-violating variant, and no fitted parameters are used in the demonstration. The main claim is defensible as a numerical methods contribution, with the caveat that the tangent/null-limiting case is not proven beyond the tested 1+1 setup.","major_comments":[],"minor_comments":[{"comment":"The abstract and Sec. II A present excision 'spacelike or tangent' to the innermost characteristic, but the tangent (null) case is implemented by the CFL-violating method II, whose convergence rests on the unproved assumption that boundary errors propagate with speeds bounded by c_- and that the error layer width converges to zero. The author's own caveat in Sec. III D is appropriate, but to avoid overclaiming, please add a qualifier in the abstract (e.g., 'spacelike excision, with numerical evidence for the null-limiting case in spherical symmetry') or explicitly flag the assumption at its first mention in Sec. II A.","section":"Sec. II A / Abstract"},{"comment":"The prescription 'excise one spatial grid point every 1/(c_- lambda) time steps' should read 1/(|c_-| lambda) (or, in practice, ceil(1/(|c_-| lambda))) since c_- is negative; the text in Sec. III D correctly uses 'every other time step' for lambda = 0.5 and |c_-| = 1.","section":"Sec. II A / Sec. III D"},{"comment":"The typesetting of the constraint equations is hard to read: 'zeta 2' should be zeta^2, and 'r zeta jr' should be r zeta j_r (with the subscript on j). Please also check the sign of the momentum constraint term against the definition j_r = -PQ, since the current notation makes it difficult to verify.","section":"Sec. III A, Eq. (9)"},{"comment":"Minor typos: 'if one uses an scheme' should be 'if one uses a scheme', and 'time lambda T' should be 'time lambda T' with a multiplication symbol or space.","section":"Sec. IV"},{"comment":"The claim of 'roughly second order convergence' would be easier to assess with a convergence-factor plot or a quantitative rate for the intermediate resolution range; currently Fig. 6 only shows the residuals at three resolutions.","section":"Fig. 6"}],"recommendation":"minor_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, modest paper. It makes the Bieri-Garfinkle-Yau expansion method into a paired excision method, derives the CFL restriction that bounds total evolution time by one light-crossing time, and shows clean second-order convergence for a self-gravitating scalar in spherical symmetry. The main result, method I, is sound. Excising one grid point per time step keeps the domain inside the domain of dependence, so no boundary conditions are needed. The convergence tests against the Einstein constraint residual are a genuine check, not just a self-consistency test.\n\nThe genuinely new pieces are the explicit framing of outer-boundary excision as boundary-condition avoidance, the CFL time-bound derivation, and the CFL-violating method II. Method I is well supported; the residual norms in Figs. 6 and 8 show second-order convergence in both the non-black-hole and black-hole cases.\n\nThe soft spot is exactly where the stress-test note points: method II excises along the null ray every 1/(c_- λ) steps, violating the CFL condition at the boundary. Convergence depends on the claim that boundary errors travel with speed bounded by c_- and stay near the boundary. The paper asserts this following Pretorius and Choptuik and checks it only in 1+1 spherical symmetry. It also cautions explicitly that other gauges and 3+1 codes may behave differently. So the concern is real but scoped. It does not touch method I, which is the core demonstration. I would not treat method II as established; I would treat it as a promising heuristic that needs a proof or a broader numerical test.\n\nThe paper is honest about limitations: the shrinking grid caps run time at about one light-crossing time, and it names applications where that is acceptable. The citation pattern is fair, including prior excision outside trapped regions [13,14] and the double-null/Bondi literature. No circularity or invented quantities.\n\nWho is this for: numerical relativists doing collapse or black-hole-interior simulations with finite difference codes. It is short, clearly written, and worth a serious referee. I would accept it for peer review; with minor revisions, mainly clarifying the method II caveat (which it already does) and possibly adding a more thorough convergence study of method II's error layer.\n\nFinal recommendation: send to a capable referee in numerical relativity. Desk rejection would be wrong; this is a useful, citable methods note.","headline":"A sound, honestly-scoped numerical-methods note: the CFL-obeying outer excision works and is demonstrated cleanly; the CFL-violating variant is a flagged conjecture, not a load-bearing part.","tokens_in":11307,"tokens_out":3180,"would_cite":true,"duration_ms":29435,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["83-08","65M06","35L60"],"pacs":[],"model":"deepseek-v4-flash","headline":"Excising a computational grid along the innermost characteristic removes the need for boundary conditions in numerical relativity.","keywords":["numerical relativity","excision method","boundary conditions","CFL condition","characteristics","Painlevé-Gullstrand coordinates","self-gravitating scalar field","spherical symmetry"],"falsifier":"Evolve a non-spherically-symmetric gravitational wave or scalar pulse in a 3+1 code using a standard gauge (BSSN or generalized harmonic) with the CFL-violating excision of Sec. III D, at three resolutions, and measure the discrete Einstein-equation residual in the interior: if the error layer grows in width or amplitude with time rather than converging to zero with resolution, the central assumption fails. The paper itself notes this test is beyond its scope.","tokens_in":10433,"feed_emoji":"🌌","tokens_out":7131,"duration_ms":66571,"temperature":0.7,"pith_summary":"This paper claims that a computational domain for a hyperbolic system such as the Einstein equations can be evolved on a compact grid with no boundary conditions at all: instead of prescribing data on an outer boundary, one excises the grid inward along a surface that is tangent to or spacelike with respect to the innermost characteristic. Since no characteristic curves enter the domain through that surface, nothing needs to be imposed there; the initial data on the original slice determine the whole evolution. The paper compares this excision method with the recently proposed expansion method and illustrates both with a code that evolves a massless self-gravitating scalar field in spherical symmetry. A key practical caveat is that with explicit finite differencing and CFL number λ≤1, a CFL-obeying excision can run only for about λ times the light-crossing time of the initial grid, though excising directly along the null ray (violating CFL at the boundary) can extend this. If the method holds in full 3+1, it offers a simpler route around the difficult problem of well-posed, constraint-preserving outer boundary conditions in numerical relativity.","feed_headline":"Excision removes boundary conditions in numerical relativity","feed_subtitle":"Shrink the grid along the innermost light ray, and initial data alone determines the evolution for about one light-crossing time.","key_machinery":"The load-bearing object is the excision surface itself: a moving outer boundary chosen so that every characteristic crossing it is outgoing rather than ingoing. Because the surface is tangent to or spacelike with respect to the innermost characteristic c−, the domain of dependence of each new time slice lies within the old grid, so one can set field values on the boundary by upwinded evolution equations instead of imposed boundary data. The machinery also includes the characteristic speed formula c± = α(±1 − ζ) in Painlevé-Gullstrand coordinates, which locates the excision surface, and the metric decomposition with a null coordinate used to describe surfaces on which no boundary conditions are needed.","core_discovery":"On the paper's own terms, the central discovery is an excision prescription: at each time step, delete the grid points whose domain of dependence is not contained in the previous time slice, choosing the excision surface so it lies tangent to or spacelike with respect to the innermost characteristic speed c−. Along such a surface no characteristics are ingoing, so the evolution is determined entirely by initial data and no boundary conditions are needed. In Painlevé-Gullstrand coordinates for spherical symmetry the characteristic speeds are c± = α(±1 − ζ), and the paper implements two variants: method I excises one grid point per time step (obeying the CFL condition, so evolution time is limited to λT), and method II excises directly along the ingoing null characteristic every 1/(c−λ) time steps (violating CFL at the boundary). Numerical tests with a self-gravitating massless scalar show roughly second-order convergence for both methods; method II develops an error layer near the excision boundary whose width shrinks to zero with resolution, while method I does not. The paper explicitly cautions that these tests are in 1+1 spherical symmetry and that the behavior of CFL-violating excision in axisymmetric or full 3+1 codes with more common gauges is not established.","pith_inferences":["A natural testable extension is to repeat the method-II comparison in axisymmetric or full 3+1 evolution with standard gauges such as generalized harmonic or BSSN; if the boundary-error layer continues to converge to zero, CFL-violating excision would become a practical tool for black-hole interior evolutions without outer-boundary prescriptions.","The same excision principle should apply to any hyperbolic system of equations, not only the Einstein equations, whenever one cares about local dynamics near an outer boundary; a scalar-wave test in flat spacetime would be a minimal check.","Combining expansion then excision could be developed into a two-phase strategy for long evolutions: expand the grid outward during the early phase, then excise during the phase where the interesting dynamics is local, avoiding both boundary-condition issues and unbounded grid growth.","If stable CFL=1 or implicit schemes are used, the run-time restriction of one light-crossing time may disappear entirely, which would make excision competitive with compactification or characteristic codes for local collapse problems."],"forward_implications":["If the excision method is correct, numerical relativity codes can evolve compact domains without solving a well-posed, constraint-preserving initial boundary value problem; initial data alone determines the solution.","For explicit finite-difference schemes with CFL number λ≤1, CFL-obeying excision restricts evolutions to about λ light-crossing times of the initial slice, making the method natural for local gravitational collapse and critical collapse rather than for extracting radiation at large distances.","With a CFL number of exactly 1, excision can move directly along the ingoing null ray while still satisfying the CFL condition, removing the run-time restriction.","The CFL-violating variant (method II) yields stable, convergent evolutions in 1+1 spherical symmetry, with any boundary-induced error confined to a layer whose thickness converges to zero with resolution.","Expansion and excision can be combined: expand the grid first, then excise once the domain reaches a desired size, permitting longer evolutions than pure excision on a fixed initial grid."],"supporting_citations":[{"why":"The expansion method this paper compares against, the recently proposed alternative that inspired the relation discussed in Secs. I and II.","marker":"[1]"},{"why":"The standard result that a boundary with no ingoing characteristics requires no boundary conditions, which grounds the excision argument.","marker":"[11]"},{"why":"A prior stable and convergent code that excises along an ingoing null ray, used to justify the error-containment assertion for CFL-violating excision.","marker":"[14]"},{"why":"The CFL stability condition that sets the λT run-time limit for explicit finite-difference schemes.","marker":"[16]"},{"why":"Earliest excision methods in numerical relativity, the tradition this method extends outside trapped regions.","marker":"[12]"},{"why":"Review of the initial-boundary-value problem challenges that motivate the search for boundary-condition-free evolution methods.","marker":"[10]"}],"fun_headline_variants":["Excision removes boundary conditions in numerical relativity","Grid excision: no boundary conditions needed for evolution","Excision method: initial data determines evolution without boundaries","Cut grid along ingoing characteristic, drop boundary conditions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"In the CFL-violating variant, the paper assumes without proof that errors introduced by violating the CFL condition at the excision boundary propagate with characteristic speeds bounded by c−, so they stay in a thin layer near the boundary that shrinks to zero with resolution; the numerical check covers only 1+1 spherical symmetry.","fun_headline_variants_meta":{"raw":{"variants":["Excision removes boundary conditions in numerical relativity","Grid excision: no boundary conditions needed for evolution","Excision method: initial data determines evolution without boundaries","Cut grid along ingoing characteristic, drop boundary conditions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0002,"raw_usage":{"total_tokens":1354,"prompt_tokens":906,"completion_tokens":448,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":522,"completion_tokens_details":{"reasoning_tokens":388}},"tokens_in":522,"tokens_out":448,"duration_ms":5091,"temperature":1.0,"reasoning_tokens":388,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:47:05.319092+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evolve a non-spherically-symmetric gravitational wave or scalar pulse in a 3+1 code using a standard gauge (BSSN or generalized harmonic) with the CFL-violating excision of Sec. III D, at three resolutions, and measure the discrete Einstein-equation residual in the interior: if the error layer grows in width or amplitude with time rather than converging to zero with resolution, the central assumption fails. The paper itself notes this test is beyond its scope.","supporting_citations":[{"cited_title":"A No-Boundary Method for Numerical Relativity","cited_arxiv_id":"1905.08657","evidence_quote":"The expansion method this paper compares against, the recently proposed alternative that inspired the relation discussed in Secs. I and II."},{"cited_title":"Kreiss and J","cited_arxiv_id":null,"evidence_quote":"The standard result that a boundary with no ingoing characteristics requires no boundary conditions, which grounds the excision argument."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"A prior stable and convergent code that excises along an ingoing null ray, used to justify the error-containment assertion for CFL-violating excision."},{"cited_title":"Seidel and W.-M","cited_arxiv_id":null,"evidence_quote":"Earliest excision methods in numerical relativity, the tradition this method extends outside trapped regions."},{"cited_title":"Sarbach and M","cited_arxiv_id":null,"evidence_quote":"Review of the initial-boundary-value problem challenges that motivate the search for boundary-condition-free evolution methods."}],"review_version":1}