{"id":"b6ed1c49-a743-4e4e-8292-90764934f50a","arxiv_id":"1908.01814","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A mixed discontinuous/continuous Galerkin scheme is proven to conserve total energy and L2 norm for general Hamiltonian evolution equations, demonstrated on Vlasov-Poisson and incompressible Euler systems.","lead":"This paper presents a new class of numerical schemes that conserve energy exactly for a wide family of physics equations written in Hamiltonian form, including the Vlasov-Poisson equations of plasma physics. The key trick is to represent the Hamiltonian in a continuous function space while the fluid or particle distribution is allowed to be discontinuous, which makes the method accurate and robust for fusion plasma simulations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Entropy monotonicity is not proven: differentiating inequality (3.15) is invalid, and central-flux DG conserves L2 but not the non-quadratic entropy functional.","rationale":"The central energy-conservation claim appears sound for the demonstrated Vlasov-Poisson and incompressible Euler cases: the w = H_h test in Proposition 3.2, the field-equation manipulation in Section 3, and the numerical tables in Section 4 are mutually consistent. The reader's Lemma 2.1 concern is real but limited: for a discontinuous Poisson tensor the proof as written does not apply, although a direct surface sum with a single-valued numerical flux for the full n*alpha*f quantity would cancel when H_h is continuous, making Remark 2.2 a missing proof rather than a false claim. The more serious soft spot is Proposition 3.4, which is internally invalid: differentiating a pointwise function inequality does not preserve the inequality for time derivatives, and central-flux DG has no reason to conserve or monotonically evolve the non-quadratic entropy functional. Since the paper explicitly advertises the entropy result in the abstract and conclusions, this should be corrected or qualified before full acceptance. The verdict remains conditional: the energy claims can stand with clarifications, but the entropy theorem needs a valid proof or removal.","tokens_in":19718,"tokens_out":14991,"duration_ms":182054,"concrete_test":"Run 1D linear advection with P1 DG, central flux, periodic boundary conditions, and smooth strictly positive initial data; integrate the semi-discrete equations with a fine SSP-RK3 step and plot S(t) = integral -f_h ln f_h. If S(t) oscillates or decreases, Proposition 3.4 is false for central fluxes. If S(t) is accidentally constant over short times, the derivative step in (3.15)-(3.16) still needs a valid proof, since the inequality-differentiation argument does not establish monotonicity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.4's proof contains an invalid step. From ln f_h <= f_h - 1 the authors obtain -f_h ln f_h >= -f_h^2 + f_h, integrate, and then differentiate both sides to conclude d/dt integral(-f_h ln f_h) >= d/dt integral(-f_h^2 + f_h). A nonnegative gap g(t) = integral[-f_h ln f_h - (-f_h^2 + f_h)] may decrease while staying nonnegative, so the derivative inequality does not follow. The issue is not merely technical: for central fluxes the semi-discrete update is skew-symmetric in the mass-weighted inner product (M f_h' = S f_h with S^T = -S), so the L2 norm is exactly conserved but non-quadratic functionals such as -f ln f generally are not conserved; the theorem would force entropy to be constant for central fluxes, which is not a consequence of the DG equations. The stated premise f_h > 0 is also not enforced by the scheme, as the paper itself notes. Thus the abstract's claim that the scheme makes entropy non-decreasing is unsupported as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a mixed discontinuous Galerkin / continuous Galerkin spatial discretization for Hamiltonian evolution equations of the form ∂f/∂t + {f,H} = 0. The distribution function is approximated in a discontinuous polynomial space V_h^p, while the Hamiltonian is constrained to the continuous subspace W_{0,h}^p = V_h^p ∩ C^0. The paper proves particle conservation, energy conservation for static Hamiltonians and for the Vlasov–Poisson total energy, and exact L2 conservation for central fluxes with monotone L2 decay for upwind fluxes. It further claims that the discrete entropy is non-decreasing under a positivity assumption. Numerical benchmarks include Vlasov–Poisson problems (free streaming, specified potentials, linear and nonlinear Landau damping) and 2D incompressible Euler flows, with references to gyrokinetic applications in the Gkeyll code.","tokens_in":19980,"tokens_out":9415,"duration_ms":105066,"significance":"The core construction is significant if the stated results hold: it gives a parameter-free spatial discretization that preserves the Hamiltonian structure's energy invariant at the semi-discrete level while retaining the option of upwinding for robustness, and it shows second-order convergence of momentum errors that is essentially independent of velocity resolution. The proofs for static Hamiltonians and for the Vlasov–Poisson system are transparent and do not rely on fitting parameters. The numerical experiments are reproducible in principle (the Gkeyll code and input files are referenced) and the convergence tests for energy and enstrophy errors confirm the expected time-stepping orders. However, two advertised claims are not established by the manuscript: the entropy monotonicity theorem and the applicability of the proofs to general non-canonical systems with discontinuous Poisson tensors, such as electromagnetic gyrokinetics. These issues are load-bearing for the abstract's generality and stability claims, so the paper needs revision before publication.","major_comments":[{"comment":"The proof of entropy monotonicity is invalid. The authors integrate the pointwise inequality −f_h ln f_h ≥ −f_h^2 + f_h and then differentiate both sides to obtain Eq. (3.16). Differentiating an inequality is not legitimate: the gap g(t) = ∫(−f_h ln f_h + f_h^2 − f_h) ≥ 0 may decrease while remaining nonnegative, so d/dt∫(−f_h^2 + f_h) ≥ 0 does not imply d/dt∫(−f_h ln f_h) ≥ 0. The issue is not merely technical: for central fluxes the semi-discrete operator conserves the mass-weighted L2 norm through a skew-symmetric structure, but that structure does not control non-quadratic functionals such as −f ln f. The manuscript itself notes that f_h > 0 is not enforced by the scheme. Therefore the abstract's claim that the scheme makes entropy non-decreasing, and the corresponding statement in Section 5, are unsupported as written. The authors should either give a valid proof (for example, a cell entropy inequality with an appropriate numerical entropy flux) or remove the entropy claim from the abstract and conclusions.","section":"§3, Proposition 3.4, Eqs. (3.15)–(3.16)"},{"comment":"The energy-conservation proof depends critically on Lemma 2.1, but the lemma is proved only under the assumption that the Poisson tensor Π_ij is continuous across cell faces. Remark 2.2 correctly notes that in electromagnetic gyrokinetics in non-orthogonal field-aligned coordinates the Poisson tensor can be discontinuous, and then proposes, without proof, that a numerical flux be used for the entire quantity n·α_h F. Consequently Proposition 3.2 and the Vlasov–Poisson energy proof do not establish energy conservation for that class of systems. The abstract's statement that the proofs apply to 'any Hamiltonian system, including ones in which the Poisson bracket operator is non-canonical (for example, the gyrokinetic equations)' is therefore broader than what the manuscript proves. Please state the continuity hypothesis on Π_ij as an explicit assumption in the theorems, prove the Remark 2.2 flux prescription, or restrict the claims to systems for which Lemma 2.1 holds.","section":"§2, Lemma 2.1 and Remark 2.2; §3, Proposition 3.2"}],"minor_comments":[{"comment":"The phrase 'quadratic invariants of total energy and the L2 norm' is imprecise: the total energy of the Vlasov–Poisson system is not a quadratic functional of f, while the L2 norm is. Suggest rewording to distinguish the L2 Casimir from the total energy.","section":"Abstract and §1"},{"comment":"There is a duplicated word: 'the flux of particles out of one cell through a particular face is is identical to the flux...' should read 'is identical'.","section":"§2, text after Eq. (2.5)"},{"comment":"The sentence 'Energy-conservation tests are performed with the same initial condition, however with on a fixed grid...' has an awkward construction and appears to omit a word; please revise.","section":"§4.1"},{"comment":"There is a typo 'updwinding' for 'upwinding' in the Conclusions, and Proposition 3.4 says 'positive definite' where 'positive' is meant for the function f_h.","section":"§5 and Proposition 3.4"},{"comment":"Reference [7] contains a typo: 'Landgon' should be 'Langdon'.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The core energy-conserving DG/CG construction is sound and worth publishing, but the advertised breadth exceeds what is proved: the entropy theorem is invalid as written and the gyrokinetic claim relies on an unproven flux prescription. A revision that fixes or removes these claims is needed before the paper can be accepted. The self-citations appear in application and context paragraphs and do not affect the central derivations, so I see no novelty-disclosure problem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my take on 1908.01814.\n\nThe paper does something real: it generalizes the Liu-Shu mixed DG/CG trick to Hamiltonian systems, proves discrete energy conservation under one clean condition (Lemma 2.1), and shows L2 conservation/decay for central/upwind fluxes. The Vlasov-Poisson and incompressible Euler benchmarks are credible, the momentum non-conservation is analyzed honestly with convergence tests, and the Gkeyll reproducibility statement is a plus. This is a solid contribution to computational plasma physics, and the scheme is already in production, which says something.\n\nThe soft spots are real but localized. The entropy claim (Prop 3.4) is not proven. The step from -f ln f >= -f^2 + f to a derivative inequality is invalid: a nonnegative gap can shrink. For central fluxes the semi-discrete operator is skew-symmetric, so L2 is conserved but non-quadratic entropies generally are not; the theorem would imply entropy is non-decreasing for central fluxes, which does not follow from the discrete system. The paper should either prove a correct entropy statement (maybe with an additional dissipation term) or drop the claim from the abstract.\n\nSecond, the abstract and conclusions overstate generality. Lemma 2.1 needs the Poisson tensor to behave so that the normal characteristic speed is continuous across faces. That holds for the applications shown, but not for every non-canonical system; Remark 2.2 admits this and suggests a flux fix without proof. The correct statement is: energy conservation holds for Hamiltonian systems satisfying Lemma 2.1, with an unproven extension for discontinuous Poisson tensors. That's still useful.\n\nLastly, the positivity caveat is stated but the entropy result is conditional on f_h > 0, which the scheme does not enforce. That is fine if the abstract says so, but currently it is buried.\n\nCitation pattern looks healthy: the central proofs don't lean on self-citations, and the Gkeyll references are in applications. No free parameters; the conservation laws are analytical.\n\nWho is this for? Anyone building conservative Eulerian kinetic or fluid solvers, especially in plasma physics. It deserves a serious referee; the core energy and L2 results are correct and important. I would send it to review and ask for a revision that fixes the entropy claim and tempers the generality statement. I would also cite it for the energy-conservation construction.","headline":"Solid energy-conserving DG extension for Hamiltonian systems, but the entropy proof is invalid and the abstract overclaims generality.","tokens_in":20456,"tokens_out":4761,"would_cite":true,"duration_ms":44798,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q83","35Q20","65M60","82D10"],"pacs":[],"model":"deepseek-v4-flash","headline":"A mixed discontinuous/continuous Galerkin discretization conserves total energy exactly for Hamiltonian evolution equations in the continuous-time limit.","keywords":["energy conservation","discontinuous Galerkin","Hamiltonian evolution equations","Vlasov-Poisson equations","incompressible Euler equations","gyrokinetics","Poisson bracket"],"falsifier":"Run the semi-discrete scheme on a two-cell Hamiltonian toy problem with a discontinuous Poisson tensor and refine both mesh and time step: exact conservation of $\\sum_j \\int_{K_j} H_h f_h\\,dz$ in the continuous-time limit must hold if the central claim generalizes, so any residual drift in this quantity would disprove it.","tokens_in":19559,"feed_emoji":"⚡","tokens_out":9162,"duration_ms":85499,"temperature":0.7,"pith_summary":"The paper sets out a spatial discretization for Hamiltonian evolution equations $\\partial f/\\partial t + \\{f,H\\}=0$ in which the distribution function is discontinuous Galerkin while the Hamiltonian is forced into the continuous subspace of the same polynomial space. It proves that, in the continuous-time limit, this mixed DG/CG choice conserves total energy exactly for any numerical flux, and that it conserves the $L_2$ norm with central fluxes or decays it monotonically with upwind fluxes. For Vlasov–Poisson the argument yields exact conservation of particle number and of total particle-plus-field energy; for any system with $f_h>0$ it yields a non-decreasing entropy. The claim matters because it gives a systematic, parameter-free rule for building stable and energy-conserving continuum schemes for plasma fluid and kinetic problems, including non-canonical systems such as gyrokinetics.","feed_headline":"Keep the Hamiltonian continuous and DG conserves energy","feed_subtitle":"A mixed DG/CG scheme for plasma and fluid equations conserves total energy with any numerical flux","key_machinery":"The load-bearing object is the pair of finite-element spaces $V^p_h$ (discontinuous piecewise polynomials for $f_h$) and $W^p_{0,h}=V^p_h\\cap C^0(Z)$ (the continuous subset used for the Hamiltonian). The mechanism is Lemma 2.1: because the characteristic velocity satisfies $\\dot{z}^i = \\Pi^{ij}\\,\\partial H/\\partial z^j$ with $\\Pi$ antisymmetric, the normal component $n_i\\dot z^i$ equals the tangential gradient of $H$ along the face, so it is continuous when $H_h$ is. This continuity cancels all inter-cell flux terms when one tests the DG equation with $H_h$, while incompressibility $\\{H_h,H_h\\}=0$ kills the volume term; the discrete field equation then supplies the field-energy half of the energy budget.","core_discovery":"The central discovery is that energy conservation is forced by a single structural choice: place the discrete Hamiltonian $H_h$ in the continuous subspace $W^p_{0,h}=V^p_h\\cap C^0(Z)$ of the discontinuous polynomial space used for $f_h$. Once $H_h$ is continuous, Lemma 2.1 shows the normal component of the characteristic velocity $\\alpha_h$ is continuous across every cell face, because that normal component is a tangential derivative of $H_h$ taken through the antisymmetric Poisson tensor. Taking the test function $w=H_h$ in the DG weak form then makes the volume term vanish through $\\{H_h,H_h\\}=0$, while the surface terms cancel exactly on summation; for Vlasov–Poisson, differentiating the discretized field equation converts the potential term into field energy, giving $d(W_k+W_E)/dt=0$. The same framework conserves particles exactly, conserves the $L_2$ norm with central flux, monotonically decays it with upwind flux, and gives an entropy inequality whenever $f_h$ stays positive.","pith_inferences":["The paper leaves implicit that the structural rule—$f$ discontinuous, $H$ in its continuous subspace—gives a general recipe: any Hamiltonian system in the form $\\partial_t f + \\{f,H\\}=0$ admits a parameter-free energy-conserving spatial discretization, so new reduced plasma models may inherit this property directly.","The unproven flux prescription in Remark 2.2 is a natural stress test: building a two-cell Hamiltonian toy problem with a jump in the Poisson tensor and checking whether total energy remains conserved would settle whether the main result extends to electromagnetic gyrokinetics in non-orthogonal coordinates.","Because the energy proof is independent of the numerical flux, a positivity-preserving limiter or reconstruction applied to $f_h$ may be compatible with energy conservation, which would repair the $f_h>0$ assumption in the entropy result.","The noted lack of a simultaneously momentum- and energy-conserving DG scheme frames a concrete open question: design a DG/CG pair that carries both invariants, or find an obstruction."],"forward_implications":["Using upwind fluxes in the DG update does not destroy exact energy conservation, so robustness and energy preservation are not in conflict at the spatial level.","The $L_2$ norm of the distribution function is conserved with central fluxes and decays monotonically with upwind fluxes, giving a nonlinear stability guarantee without added dissipation.","For Vlasov–Poisson, total particle-plus-field energy is conserved in continuous time; with a Runge–Kutta time stepper the only energy error is the time-stepping error, which converges at the RK order.","Momentum is not conserved exactly, but momentum error converges with spatial resolution alone, so coarse velocity grids do not worsen it.","The same construction applies to non-canonical Poisson brackets such as gyrokinetic equations whenever the normal characteristic velocity is continuous."],"supporting_citations":[{"why":"Supplies the original DG method for 2D incompressible flows, including the enstrophy conservation/decay proof the present scheme generalizes.","marker":"[35]"},{"why":"Supplies the Hamiltonian and Poisson-bracket formalism, including the Poisson tensor used in Lemma 2.1.","marker":"[10]"},{"why":"Provides an earlier energy-conserving DG method for Vlasov–Poisson, used as the contrast case for cost and polynomial-degree restrictions.","marker":"[3]"},{"why":"Establishes the energy-conserving DG approach for Vlasov–Maxwell and supplies the warning that under-integration can break energy conservation.","marker":"[31]"},{"why":"Carries the discrete total-energy proof to the full-f electromagnetic gyrokinetic system, showing the scheme extends beyond the benchmark examples.","marker":"[37]"}],"fun_headline_variants":["Continuous Hamiltonian: key to DG energy conservation","Make H continuous, DG conserves energy exactly","One structural choice gives DG energy conservation","Mixed DG/CG with continuous H: energy is conserved"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 2.1, which presumes the Poisson tensor is smooth enough that the normal component of the phase-space velocity is continuous across cell faces; the entropy result additionally presumes $f_h>0$, which the scheme itself does not enforce.","fun_headline_variants_meta":{"raw":{"variants":["Continuous Hamiltonian: key to DG energy conservation","Make H continuous, DG conserves energy exactly","One structural choice gives DG energy conservation","Mixed DG/CG with continuous H: energy is conserved"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000261,"raw_usage":{"total_tokens":1660,"prompt_tokens":1081,"completion_tokens":579,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":697,"completion_tokens_details":{"reasoning_tokens":521}},"tokens_in":697,"tokens_out":579,"duration_ms":5975,"temperature":1.0,"reasoning_tokens":521,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:01:27.659899+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the semi-discrete scheme on a two-cell Hamiltonian toy problem with a discontinuous Poisson tensor and refine both mesh and time step: exact conservation of $\\sum_j \\int_{K_j} H_h f_h\\,dz$ in the continuous-time limit must hold if the central claim generalizes, so any residual drift in this quantity would disprove it.","supporting_citations":[{"cited_title":"Liu and C.-W","cited_arxiv_id":null,"evidence_quote":"Supplies the original DG method for 2D incompressible flows, including the enstrophy conservation/decay proof the present scheme generalizes."},{"cited_title":"Cary and A","cited_arxiv_id":null,"evidence_quote":"Supplies the Hamiltonian and Poisson-bracket formalism, including the Poisson tensor used in Lemma 2.1."},{"cited_title":"Ayuso, J","cited_arxiv_id":null,"evidence_quote":"Provides an earlier energy-conserving DG method for Vlasov–Poisson, used as the contrast case for cost and polynomial-degree restrictions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Carries the discrete total-energy proof to the full-f electromagnetic gyrokinetic system, showing the scheme extends beyond the benchmark examples."}],"review_version":1}