{"id":"a8ab5118-78a7-483c-926f-0b0b126a514e","arxiv_id":"2505.07304","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors give the first degree bounds for closure properties (sum, product, quotient, composition) of D-algebraic functions, under a complete-intersection genericness condition.","lead":"This paper proves upper bounds on the size of polynomial differential equations obtained by adding, multiplying, dividing, or composing D-algebraic functions, which are functions defined by nonlinear differential equations. The bounds grow fast and depend on a technical 'generic position' condition on the input equations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"D-regularity is an unproven genericity hypothesis: Example 11 shows it fails for natural D-algebraic systems, so the advertised closure bounds only hold conditionally.","rationale":"The main theorem's proof is largely sound: despite the incorrect wording in Proposition 7, the final Hilbert-function bound d^{ρ+1} binom(r_min+k,k) survives when the homogenization variable is counted correctly, and the dimension comparison leading to the stated threshold is valid. The corollaries for sum/product and composition rely on the same D-regularity condition and on proof sketches that can be filled in. The genuine load-bearing limitation is the unproven genericity of D-regularity: the paper's own Example 11 shows a D-algebraic ideal with a natural presentation that is not D-regular, and no theorem guarantees a D-regular presentation for closure outputs. The authors are honest about this limitation, but the advertised 'first degree bound for closure operations' is therefore conditional on a hypothesis whose prevalence is only supported by unreported experiments. Since the reader's verdict is already CONDITIONAL and identifies exactly this assumption, my stress test does not move the verdict; it sharpens the reason for the condition.","tokens_in":15044,"tokens_out":34526,"duration_ms":358916,"concrete_test":"Take n=2, r1=r2=2, and 100 random triples (P1,P2,Q) with P1 ∈ K[y1,y1',y1''], P2 ∈ K[y2,y2',y2''] of total degree 3, and Q ∈ K[y1,y1',y1'',y2,y2',y2''] of total degree 2. For each, form I = ⟨P1,P2,z−Q⟩ and check whether the homogenized derivative tuple h(P1)^{(0..2)}, h(P2)^{(0..2)}, h(z−Q)^{(0..2)} is a complete intersection by computing the Hilbert series of the generated ideal and comparing it with the regular-sequence series ∏(1−t^{d_i})^3/(1−t)^N. Also rerun Example 11 to confirm that the original P1,P2 fail this test while Q1,Q2 pass. If a non-negligible fraction of random closure systems fail, the paper's claim that D-regularity 'often holds' is unsupported; if all pass, the remaining risk is the missing proof for arbitrary inputs.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim depends on D-regularity (Definition 10), which requires the homogenized derivative tuple (h(P_j)^{(k)})_{1≤j≤n,0≤k≤ρ} to be a complete intersection. The paper never shows that an arbitrary D-algebraic system can be transformed into a D-regular one, nor that the systems produced by the closure algorithms of [1] satisfy this condition. Example 11 is an explicit D-algebraic system in which f2 is a sum of two D-algebraic functions and the original tuple is not D-regular: I^{(2)} ∩ Q[y2,y2',...] = {0}, and D-regularity is recovered only by replacing P1,P2 with ad hoc Q1,Q2 that use closed-form knowledge of the solutions. The only evidence offered for the hypothesis is the sentence that experiments 'seem to indicate that this hypothesis is often satisfied,' with no data or reproducible protocol. Consequently the advertised 'first degree bound for closure operations' is not established for arbitrary D-algebraic inputs; its scope is an unproven genericity condition that is known to fail on a natural example.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript gives degree bounds for the polynomial differential equations obtained from D-algebraic closure properties (addition, multiplication, division, composition). The authors define a D-regularity condition on a tuple (P_1,...,P_n), requiring that the homogenized derivatives h(P_j)^{(k)} form a complete intersection, and prove (Theorem 12) that under this hypothesis the elimination ideal contains a nonzero element of prescribed order r and degree bounded by an expression in the input degrees and orders. The paper then derives corollaries for algebraic operations and composition, and gives separate resultant-based degree bounds for eliminating algebraic functions, hyperexponential functions, and the independent variable x. The authors explicitly acknowledge that D-regularity is nontrivial; Example 11 illustrates a natural D-algebraic system where it fails.","tokens_in":15236,"tokens_out":49404,"duration_ms":507144,"significance":"The paper addresses a natural open question: for D-finite closures both order and degree bounds are known, whereas for D-algebraic closures only order bounds were previously available. If the main bound is correct, the paper provides the first general degree bounds for closure operations on D-algebraic functions. The proof is self-contained, uses standard Hilbert-function and resultant techniques, and contains no fitted parameters. The main limitation is that the advertised bounds are conditional on D-regularity, a hypothesis that is not characterized, is not shown to hold for the systems produced by the known closure algorithms, and can fail on natural examples. The resultant-based bounds of Section 7 are unconditional and are a useful contribution.","major_comments":[{"comment":"The stated Hilbert series formula is incorrect as written. The ring R=K[s,x_1,...,x_n] has n+1 variables, so its Hilbert series is (1-t)^{-(n+1)}, not (1-t)^{-n}. For a regular sequence of k homogeneous polynomials, the denominator should be (1-t)^{n+1}. This false identity is later used in the proof of Theorem 12 and should be corrected.","section":"Section 2, Proposition 7"},{"comment":"The displayed Hilbert series for I^{(r-r_l)} has denominator exponent r_min+n(r-r_l+1), but the ambient ring R^h_{r+r_1,...,r+r_n} has one more variable, so the correct denominator exponent is r_min+n(r-r_l+1)+1. The subsequent coefficient bound HF ≤ d^{r-r_l+1} binom(r_min+k,k) can still be justified for the corrected series by the same induction argument, but the manuscript should state this explicitly. As written, the proof of the main theorem contains a false intermediate identity.","section":"Section 4, proof of Theorem 12"},{"comment":"The bound HF_{h(I)}(k) ≤ (r_1+r_2+1)! d_1^{r_2} d_2^{r_1} binom(r_1+r_2+k, r_1+r_2) is stated without derivation. A direct application of the coefficient lemma used in Theorem 12 would give additional factors d_1 d_2, so a sharper coefficient estimate is needed. In addition, the claim that the homogenized family (P_1^{(j)}, P_2^{(j)}, d^i(z-y_1)) is a complete intersection is only asserted by analogy with Proposition 13; the dimension argument should be written out.","section":"Section 6, proof of Proposition 16"},{"comment":"The main theorems are conditional on D-regularity, but the paper gives no criterion for verifying this condition and does not prove that the systems produced by the closure algorithms of [1] are D-regular. Example 11 shows that a natural D-algebraic system arising from a sum need not be D-regular. Since the title and introduction present the contribution as bounds for closure properties, the authors should state the conditional scope precisely in the abstract and introduction, and discuss whether D-regularity can be checked or enforced.","section":"Sections 4 and 5, scope of D-regularity"}],"minor_comments":[{"comment":"The proof states that a zero divisor modulo I_i must lie in a minimal prime ideal; this is only immediate after noting that complete intersection ideals are unmixed, and the missing justification should be supplied or replaced by a reference.","section":"Section 2, proof of Proposition 8"},{"comment":"The theorem statement should read 'suppose that (P_1,...,P_n) is D-regular at order r-r_l', and the phrase 'degree k' should be clarified to mean total degree at most k.","section":"Section 4, Theorem 12"},{"comment":"The ambient polynomial ring in which the homogenized derivative tuple is required to be a complete intersection should be specified explicitly.","section":"Section 4, Definition 10"},{"comment":"The caption contains 'r □ rmin'; this should read 'r − r_min'.","section":"Figure 1 caption"},{"comment":"Reference [5] contains the typo 'laballed trees'; it should be 'labelled trees'.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper has a sound core idea and the resultant-based special cases are useful. The central proof, however, contains a concrete error in the Hilbert series formula for regular sequences, and the same issue affects the counting in Proposition 16. The final bound in Theorem 12 may still be recoverable by proving the coefficient bound directly for the corrected series, but the manuscript as submitted asserts a false identity in a load-bearing position. I would also ask the authors to be more explicit about the scope of the D-regularity hypothesis and, if the experimental claim is kept, to provide the protocol behind the statement that the hypothesis is 'often satisfied'."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read through 2505.07304 with some care. The real news is that it gives degree bounds for closure properties of D-algebraic functions, where only order bounds were known before. The Hilbert-function angle is new, and the resultant-based special cases (Prop 17–19) are clean, elementary, and I believe correct. The authors also deserve credit for being upfront about the size of the bounds and for admitting they couldn't test the main theorem directly.\n\nThe soft spots, in order of importance. First, D-regularity is a loaded assumption. Theorem 12 only fires when the homogenized derivative tuple is a complete intersection, and the paper does not show that natural D-algebraic systems, or the systems produced by the closure algorithms of [1], actually satisfy this. Example 11 exhibits a simple D-algebraic system where the original tuple is not D-regular, and the recovery relies on knowing closed-form solutions. The only support is a sentence about experiments \"seem to indicate\"—no data, no protocol. So the advertised \"first degree bound\" is conditional on a genericity hypothesis that is known to fail on a natural example. This is the main thing holding the paper back.\n\nSecond, there is a concrete mathematical error in the written proof. Proposition 7 states that the Hilbert series of K[s,x1,...,xn] is (1-t)^(-n)—that's wrong, it should be (1-t)^(-(n+1)). This formula is used in the proof of Theorem 12, so the exact bound as written is not supported. My guess is it's a systematic off-by-one that can be repaired without changing the qualitative conclusion, but as submitted the proof is not correct.\n\nThird, Proposition 16 reads as a sketch; the factorial factor in the bound is crude and the complete-intersection claim is asserted rather than shown. Minor relative to the above.\n\nBottom line: this is a solid conditional result, and the flaws look fixable. It should go to a serious referee. I'd ask the authors to fix the Hilbert-series error, give a more careful treatment of D-regularity (what classes of inputs guarantee it, or how to detect it), and flesh out Prop 16. If you work on D-algebraic closure, the resultant bounds alone are worth citing.","headline":"Gives the first degree bounds for D-algebraic closure properties, but the advertised bounds hinge on a D-regularity condition that is not shown to be generic, and the written proof contains a repairable Hilbert-series error.","tokens_in":15749,"tokens_out":3414,"would_cite":true,"duration_ms":32888,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["12H05","13P10","68W30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves explicit degree bounds for the polynomial differential equations obtained by closure operations on D-algebraic functions, under a complete-intersection condition on the defining equations.","keywords":["D-algebraic functions","degree bounds","differential elimination","complete intersection","Hilbert series","closure properties","order-degree curves","differential algebra"],"falsifier":"Run a complete-intersection test on the homogenized derivative systems of many random D-algebraic inputs of order 2 or 3; because Example 11 already exhibits a D-algebraic system that is not D-regular, the rate of failure is the quantity that decides how widely the bounds apply.","tokens_in":14823,"feed_emoji":"🧮","tokens_out":9243,"duration_ms":89627,"temperature":0.7,"pith_summary":"D-algebraic functions, meaning solutions of nonlinear polynomial differential equations, are closed under addition, multiplication, division, and composition. Until now, algorithms for these closure operations could bound the order of the resulting equation but not its degree. This paper proves the first degree bounds, under a genericity condition called D-regularity: when the homogenized derivatives of the defining polynomials form a complete intersection, the elimination ideal must contain a nonzero equation whose degree does not exceed an explicit expression in the input degrees and orders. The bounds are large, and the paper argues this is not mere slack: generic closure can genuinely produce equations that are too big to write out. Sharper resultant-based bounds are proved for the special cases of eliminating algebraic or hyperexponential functions and the variable $x$.","feed_headline":"First degree bounds for D-algebraic closure equations","feed_subtitle":"Under a generic complete-intersection condition, sums, products, and compositions of D-algebraic functions get concrete size bounds.","key_machinery":"The carrying object is the tuple of homogenized derivatives $(h(P_j)^{(k)})_{1\\le j\\le n,\\,0\\le k\\le \\rho}$; D-regularity means exactly that this tuple is a complete intersection in the homogenized polynomial ring. For such a tuple the Hilbert series of the ideal it generates is known explicitly, $\\prod_i (1-t^{d_i})^{\\rho+1}/(1-t)^{r_{\\min}+n(\\rho+1)}$. The proof compares the coefficient of $t^k$ with the dimension of the space of degree-$k$ polynomials in the variables $s,y_l,\\dots,y_l^{(r)}$; once the former is smaller than the latter, some degree-$k$ polynomial must map to zero in the quotient, and setting $s=1$ produces the desired element of the elimination ideal. The resultant arguments in the special cases replace this counting by Sylvester-matrix degree estimates.","core_discovery":"On its own terms, the paper establishes a conditional size bound for differential elimination. Given differential polynomials $P_1,\\dots,P_n$ that define D-algebraic functions and satisfy a D-regularity condition, the elimination ideal $\\langle P_1,\\dots,P_n\\rangle^{(r-r_l)}\\cap K[y_l,y'_l,\\dots]$ is shown to contain a nonzero element of order $r$ as soon as its degree $k$ exceeds $(r+1)(d^{1+(r_{\\min}-r_l)/(r-r_{\\min}+1)}-1)$, where $d$ is the product of the degrees of the $P_j$ and $r_{\\min}$ is the sum of their orders. From this, the paper derives degree bounds for equations satisfied by $Q(f_1,\\dots,f_n)$, covering addition and multiplication, for quotients under an additional joint D-regularity hypothesis, and for compositions, where the bound involves $(r_1+r_2+1)!\\,d_2^{r_1}d_1^{r_2}$. It also gives sharper resultant-based bounds for the special cases of eliminating an algebraic function, a hyperexponential function, or the independent variable $x$.","pith_inferences":["If the experimental indication that D-regularity is generic holds, the bound in Theorem 12 is a realistic worst-case estimate; the paper's own comparison with algebraic dependence suggests the counting argument may overshoot by a factor of $r+1$, so a tighter count would immediately improve the bound.","The same Hilbert-series device should transfer to differential ideals in several variables or to partial differential equations, where a complete-intersection condition on prolongations would play the same role.","The order-degree phenomenon visible in the bound implies a practical algorithmic heuristic: when looking for a defining equation of a D-algebraic combination, trying a few extra derivative orders before raising the target degree can reduce the number of monomials that must be handled.","The resultant-based cases indicate that structure-specific elimination can beat the generic bound by orders of magnitude; identifying more such structures is a natural extension."],"forward_implications":["For every D-regular tuple, a nonzero differential equation for the selected component of order $r$ exists with degree at most the stated threshold; this is the first degree bound for outputs of D-algebraic closure operations.","Sums, products, and more generally $Q(f_1,\\dots,f_n)$ inherit the bound whenever each input equation is D-regular, while quotients require a joint D-regularity hypothesis on $(P_1,\\dots,P_n,Q_d z-Q_n)$.","Compositions satisfy an equation of order $r_1+r_2$ and degree bounded by $(r_1+r_2+1)((r_1+r_2+1)!\\,d_2^{r_1}d_1^{r_2}-1)$, and the same equation works for every valid composition.","The size of the bounds supports the paper's explanation for why some combinatorial D-algebraic functions have no explicitly written defining equation: generic closure can make the equation prohibitively large.","For eliminating an algebraic function, a hyperexponential function, or the variable $x$, resultant-based degree bounds are linear in the input degrees rather than exponential."],"supporting_citations":[{"why":"Supplies the closure-property algorithms and the motivating examples, including the system used to show D-regularity can fail, whose output size this paper bounds.","marker":"[1]"},{"why":"Source for the Hilbert series, Hilbert polynomial, and complete-intersection facts used in the proof of Theorem 12.","marker":"[13]"},{"why":"Quoted for the dimension theorem used in Proposition 8 to equate complete intersections with regular sequences.","marker":"[24]"},{"why":"Provides the analogous degree bounds for D-finite closure properties that this paper extends to the D-algebraic setting.","marker":"[16]"},{"why":"Standard reference for D-finite functions and the combination of order and degree bounds that motivates the present questions.","marker":"[17]"},{"why":"Introduces order-degree curves for linear operators, against which the paper's nonlinear analogue is compared.","marker":"[11]"},{"why":"Cited as giving better bounds in more specific situations, marking the contrast with the generic bounds proved here.","marker":"[19]"}],"fun_headline_variants":["Conditional degree bounds for D-algebraic closure operations","Degree bounds for D-algebraic sums, products, and compositions","Explicit size bounds for D-algebraic closure under a condition","First degree bounds for D-algebraic elimination under regularity","Bounds on polynomial differential equations for D-algebraic functions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The degree bounds all rest on the assumption that the homogenized derivatives of the defining equations form a complete intersection, the D-regular condition, and the paper does not prove this condition must hold for every D-algebraic system; Example 11 shows it can fail.","fun_headline_variants_meta":{"raw":{"variants":["Conditional degree bounds for D-algebraic closure operations","Degree bounds for D-algebraic sums, products, and compositions","Explicit size bounds for D-algebraic closure under a condition","First degree bounds for D-algebraic elimination under regularity","Bounds on polynomial differential equations for D-algebraic functions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000865,"raw_usage":{"total_tokens":3689,"prompt_tokens":824,"completion_tokens":2865,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":2780}},"tokens_in":440,"tokens_out":2865,"duration_ms":21434,"temperature":1.0,"reasoning_tokens":2780,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T22:22:45.404831+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a complete-intersection test on the homogenized derivative systems of many random D-algebraic inputs of order 2 or 3; because Example 11 already exhibits a D-algebraic system that is not D-regular, the rate of failure is the quantity that decides how widely the bounds apply.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the closure-property algorithms and the motivating examples, including the system used to show D-regularity can fail, whose output size this paper bounds."},{"cited_title":"Shafarevich and Miles Reid","cited_arxiv_id":null,"evidence_quote":"Quoted for the dimension theorem used in Proposition 8 to equate complete intersections with regular sequences."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the analogous degree bounds for D-finite closure properties that this paper extends to the D-algebraic setting."},{"cited_title":"2023.D-Finite Functions","cited_arxiv_id":null,"evidence_quote":"Standard reference for D-finite functions and the combination of order and degree bounds that motivates the present questions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces order-degree curves for linear operators, against which the paper's nonlinear analogue is compared."}],"review_version":1}