{"id":"0070a226-0cf4-41ac-a528-eec5725a6605","arxiv_id":"2411.13526","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"CM elliptic curves over Q have natural density zero among all elliptic curves ordered by naive height, and the j=0 family accounts for 100% of the CM curves asymptotically.","lead":"Elliptic curves are mathematical objects that appear throughout number theory. This paper counts how many of them have complex multiplication, a special hidden symmetry, when the curves are ordered by a standard size measure.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The paper's central claims are (1) density zero of CM curves in the family E and (2) 100% of CM curves have j=0. Both follow from matching exponents: #E(X)=c X^{5/6}, #E_0(X)=c X^{1/2}, #E_1728(X)=c X^{1/3}, and #E_j(X)=O(X^{1/6}) for the other eleven j-invariants. I examined the only potentially delicate input, Lemma 2.3's bound on integral points on cuspidal cubics. The parametrization t ↦ (t^2/a, t^3/a) is a bijection, and integrality forces the denominator s of t in lowest terms to divide q, so for each divisor s of q the numerator r is confined to an interval of length O((pqT)^{1/2}). This is a correct elementary argument and yields the claimed O(T^{1/2}). Even with the slight algebraic slip in the displayed coefficient in Eq. (2.25), the corrected coefficient remains nonzero and fixed for each of the eleven CM j-invariants, so the O(X^{1/6}) bound is unaffected. The Brumer count for #E(X) is reproduced with the chosen height normalization and the singular-curve error is O(X^{1/6}), smaller than the main term. The class-number-one classification is standard. I therefore see no load-bearing concern; the reader's ACCEPT verdict stands unchanged.","tokens_in":19418,"tokens_out":30585,"duration_ms":322445,"concrete_test":"Re-derive Eq. (2.25) from j=1728*4A^3/(4A^3+27B^2): the correct relation is B^2=4(1728-j)/(27j) A^3. Verify that for each j in J_cm \\ {0,1728} the corrected rational a_j is nonzero, and rerun the proof of Theorem 2.4 with a_j replaced accordingly. If any j gave a_j=0, or if the integral-point count exceeded O(X^{1/6}), the claim would be threatened; neither occurs.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After re-checking the argument, I find no load-bearing flaw in the central claims. The most delicate step is Lemma 2.3, and it is sound: for a=p/q in lowest terms, integral points on y^2=a x^3 with |x|≤T correspond to t=r/s in lowest terms with s^2|q and |r|≤(pqT)^{1/2}, giving O(T^{1/2}) with constant depending on a. Applied with T=X^{1/3}/2^{2/3}, this gives #E_j(X)=O(X^{1/6}) for each j≠0,1728, which is far smaller than the X^{1/2} main term from j=0, so Theorem 1.5 follows. One minor algebraic slip: the coefficient in Eq. (2.25) should be 4(1728-j)/(27j), not (1728-4j)/(27j). This does not change the exponent in the bound, since the corrected coefficient is still a fixed nonzero rational for the eleven CM j-invariants, so Lemma 2.3 still applies with the same O(X^{1/6}). No missing support or circular step was found.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the density and distribution of CM elliptic curves over Q in the family E of minimal short Weierstrass models, ordered by the naive height h = max{4|A|^3, 27|B|^2}. The main results are: (1) the natural density of CM curves in E is zero (Theorem 1.1); (2) among CM curves, those with j-invariant 0 have asymptotic density 1 (Theorem 1.5); (3) for j=0 and j=1728, precise asymptotic counts are obtained, and for the other eleven CM j-invariants an upper bound O(X^{1/6}) is proved (Theorem 1.6/2.4); and (4) an alternative family of representatives built from twists is analyzed, yielding leading terms for every CM j-invariant (Theorem 4.4). The proofs are elementary, using k-free counts and a lattice-point estimate on cuspidal cubics.","tokens_in":19620,"tokens_out":34174,"duration_ms":306243,"significance":"If correct, the paper establishes a striking and previously unquantified phenomenon: although infinitely many Q-isomorphism classes of CM elliptic curves exist for each of the thirteen class-number-one orders, the j=0 family absorbs all the mass when curves are ordered by naive height, and CM curves are negligibly rare among all elliptic curves. The arguments are internally consistent and free of fitted parameters; the leading constants are exact expressions in zeta values. A particular strength is that the main load-bearing estimate (Lemma 2.3 on integral points on cuspidal cubics) is proved in detail and is sound. The paper also provides reproducible code, which is a plus. The claimed results follow from the stated lemmas, and I found no circular reasoning.","major_comments":[],"minor_comments":[{"comment":"In the proof of Theorem 2.4, solving j = 1728 * 4A^3 / (4A^3 + 27B^2) for B^2 gives B^2 = 4(1728-j)/(27j) A^3, so the coefficient displayed in Eq. (2.25) should be 4(1728-j)/(27j), not (1728-4j)/(27j). The error does not affect the subsequent O(X^{1/6}) bound because the corrected constant is still a fixed nonzero rational for the eleven j-invariants in question, but the displayed equation should be corrected.","section":"§2, Eq. (2.25)"},{"comment":"In the line following Eq. (2.24), the simplification 2/(2^{2/3}ζ(4))X^{1/3} = 2^{1/2}/ζ(4) X^{1/3} is incorrect; the correct exponent is 2^{1/3}, consistent with the theorem statement. This is a typographical slip only.","section":"§2, proof of Theorem 2.4, Eq. (2.24)"},{"comment":"The explicit constant in Eq. (2.10), 2σ0(q)√|pq| T^{1/2}, is not valid uniformly for every T>0 because the count of integers in a bounded interval gives 2N+1 rather than 2N; the O(T^{1/2}) statement in Eq. (2.11) is correct, but the lemma should restrict to T ≥ 1 or use a slightly larger constant.","section":"§2, Lemma 2.3"},{"comment":"The row for dK=-7, f=1 (j=-3375) in Table 6 reports ET_j(10^10)=0, which is inconsistent with Table 4 (E_j(10^10)=8) and with Theorem 4.4 using the base curve y^2=x^3-35x+98 of height 259308, which would predict 8 twists. This entry should be checked and corrected.","section":"§4, Table 6"},{"comment":"The decomposition D'(X) = ⨆_{d≤X^{1/12}} d * M(d^{-12}X) is correct, but the uniqueness of the representation is not justified in the text; a one-sentence explanation would help the reader.","section":"§2, proof of Theorem 2.1, Eq. (2.2)"},{"comment":"The running title contains the typo 'CUR VES'; it should be 'CURVES'.","section":"Title"}],"recommendation":"minor_revision","confidential_remarks":"The core mathematics is sound and the proofs are internally consistent. The revisions needed are local: fix the algebraic typo in Eq. (2.25), the simplification typo around Eq. (2.24), the uniformity statement in Lemma 2.3, and the apparent inconsistency in Table 6 for j=-3375. I would be happy to see the revised version published."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nShort version: this is a serious paper and the main claims hold. The authors prove that CM elliptic curves over Q have natural density zero in the standard minimal Weierstrass family ordered by naive height, and that among CM curves, 100% asymptotically have j=0. The explicit asymptotic for #E_cm(X) — with constants 2/(3^{3/2}ζ(6)) X^{1/2} + 2^{1/3}/ζ(4) X^{1/3} — and the fixed-j counts are not in Brumer or anywhere else I know.\n\nWhat it does well: the counting is clean and self-contained. The j=0 and j=1728 cases reduce to k-free integer counts with zeta constants; the other eleven j-invariants are handled by a lattice-point bound on y^2 = a x^3. The error terms are tracked, no fitted constants, no circularity. The density-zero result and the 100% j=0 statement follow directly from those asymptotics. Section 4 redoes the count in a twist-based family and gets a genuine X^{1/6} leading term for j≠0,1728, which is a nice complement.\n\nSoft spots, all minor. There is a slip in equation (2.25): the coefficient should be 4(1728-j)/(27j), not (1728-4j)/(27j). Since j is fixed and nonzero for the eleven CM values, the exponent in the O(X^{1/6}) bound is unaffected, so no repair is needed beyond fixing the typo. Section 4's constants C(j) depend on the chosen representative E_j; the authors acknowledge this only in a remark, and a careful statement about which quantities are representation-independent would help. The code is provided but not versioned; a commit hash would make the numerics reproducible. The numerics themselves match the main terms well.\n\nOne thing I pushed on: the stress-test concern about Lemma 2.3 being load-bearing. It is sound, and the parametrization argument is correct. The 100% j=0 conclusion really does rest on that O(X^{1/6}) bound for the other eleven j-invariants, but the bound is proved properly, so I don't consider it a hidden assumption.\n\nWho should read this: arithmetic geometers and number theorists interested in CM curves or average ranks. It is a modest but real contribution, and the 100% j=0 fact is likely to be cited. Verdict: accept, with the typo fixed and representative-dependence made explicit. I would send it to a serious referee; it deserves one.","headline":"A clean, correct counting argument that turns the folklore heuristic into a theorem; send to a serious referee.","tokens_in":20186,"tokens_out":2181,"would_cite":true,"duration_ms":22477,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G05","11G15","11N45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that CM elliptic curves over $\\mathbb{Q}$ have natural density 0 in the family of all elliptic curves ordered by naive height, and among CM curves, those with $j$-invariant 0 have natural density 1.","keywords":["elliptic curves","complex multiplication","natural density","naive height","j-invariant","CM orders of class number one","integral points on cuspidal cubics","power-free integers"],"falsifier":"Compute $\\#\\mathcal{E}_j(X)$ for one fixed CM $j$-invariant $j\\neq 0,1728$, such as $j=-3375$, up to $X=10^{12}$ using the counting code described in the paper; if the count grows faster than a constant multiple of $X^{1/6}$ (for instance like $X^{1/3}$ or $X^{1/2}$), the pivotal Lemma 2.3 bound fails and the 100% $j=0$ density conclusion would not follow. A more direct check is to search for integral points on $y^2=\\frac{1728-4j}{27j}x^3$ with $|x|\\le T$ and confirm whether their number exceeds the bound $2\\sigma_0(q)\\sqrt{|pq|}\\,T^{1/2}$ from the lemma.","tokens_in":19213,"feed_emoji":"📉","tokens_out":16422,"duration_ms":134346,"temperature":0.7,"pith_summary":"Complex multiplication is rare among elliptic curves over $\\mathbb{Q}$: the paper proves that when all curves $y^2=x^3+Ax+B$ in minimal short Weierstrass form are ordered by the naive height $h_{\\mathrm{naive}}=\\max\\{4|A|^3,27|B|^2\\}$, the CM curves have natural density $0$. It then proves that among the thirteen CM orders of class number one, the curves with $j$-invariant $0$ have natural density $1$ inside the CM family, so asymptotically all CM curves over $\\mathbb{Q}$ are the $j=0$ curves with CM by $\\mathbb{Z}[( -1+\\sqrt{-3})/2]$. The proof gives asymptotic counts for the number of curves with a fixed $j$-invariant: the $j=0$ family contributes a main term of order $X^{1/2}$, the $j=1728$ family of order $X^{1/3}$, and each of the other eleven CM $j$-invariants at most $O(X^{1/6})$, against a total of order $X^{5/6}$. These exponents force both density statements, and the same 100% conclusion is obtained in an alternative family of representatives built from the theory of twists.","feed_headline":"Density zero for CM curves; j=0 takes all the mass","feed_subtitle":"Ordered by naive height, CM curves vanish; the j=0 curves take all of them asymptotically.","key_machinery":"The central mechanism is counting integral points on cuspidal cubics. For a fixed $j\\neq 0,1728$, the condition $j(E_{A,B})=j$ is equivalent to the point $(A,B)$ lying on the curve $y^2=\\frac{1728-4j}{27j}x^3$. Lemma 2.3 shows that any rational cuspidal cubic $y^2=a x^3$ with $a=p/q$ in lowest terms has at most $2\\sigma_0(q)\\sqrt{|pq|}\\,T^{1/2}$ integral points with $|x|\\le T$; the proof parametrizes the curve by $t\\mapsto (t^2/a,t^3/a)$ and observes that integrality forces the denominator of $t$ to divide $q$. Applied with $a=(1728-4j)/(27j)$ and $T=X^{1/3}/2^{2/3}$, this bounds each of the eleven non-exceptional CM $j$-invariants by $O(X^{1/6})$. For $j=0$ and $j=1728$, the curves are forced to have $A=0$ or $B=0$, respectively, so the count reduces to counting $6$-th and $4$-th power-free integers, giving the $X^{1/2}$ and $X^{1/3}$ main terms. The total $\\#\\mathcal{E}(X)$ is Brumer's lattice-point count of order $X^{5/6}$, and the thirteen CM $j$-invariants come from the classification of orders of class number one.","core_discovery":"The paper establishes that the natural density of the set $\\mathcal{E}^{\\mathrm{cm}}$ of CM elliptic curves inside the family $\\mathcal{E}$ of minimal short Weierstrass models over $\\mathbb{Q}$ is $0$, and that the natural density of the subfamily $\\mathcal{E}_0$ with $j$-invariant $0$ inside $\\mathcal{E}^{\\mathrm{cm}}$ is $1$. In concrete terms, asymptotically none of the elliptic curves over $\\mathbb{Q}$, ordered by naive height, have complex multiplication, and asymptotically all of the CM curves that do occur have $j=0$. This follows from the asymptotic formulas $\\#\\mathcal{E}_0(X)=\\frac{2}{3^{3/2}\\zeta(6)}X^{1/2}+O(X^{1/12})$, $\\#\\mathcal{E}_{1728}(X)=\\frac{2^{1/3}}{\\zeta(4)}X^{1/3}+O(X^{1/12})$, $\\#\\mathcal{E}_j(X)=O(X^{1/6})$ for the other eleven CM $j$-invariants, and Brumer's total $\\#\\mathcal{E}(X)=\\frac{2^{4/3}}{3^{3/2}\\zeta(10)}X^{5/6}+O(X^{7/12})$. Combining these gives $\\#\\mathcal{E}^{\\mathrm{cm}}(X)=\\frac{2}{3^{3/2}\\zeta(6)}X^{1/2}+\\frac{2^{1/3}}{\\zeta(4)}X^{1/3}+O(X^{1/6})$, so the CM subfamily grows like $X^{1/2}$ inside an $X^{5/6}$ family and the $j=0$ subfamily dominates the CM subfamily at the density level.","pith_inferences":["Since the proof gives only an $O(X^{1/6})$ bound in the minimal-Weierstrass family for the eleven non-exceptional CM $j$-invariants, while the twist family yields an explicit main term, a natural next step is to determine the true asymptotic for $\\#\\mathcal{E}_j(X)$ for these $j$; the constants $C(j)$ from the twist family provide a candidate for the leading coefficient.","The density conclusions are tied to the box-shaped height balls defined by $h_{\\mathrm{naive}}$; under other standard orderings, such as by conductor or Faltings height, the relative growth of CM curves could differ, since the lattice-point mechanism depends on the precise box geometry.","The twist-based family $\\mathcal{E}_T$ separates the $j$-invariant from a power-free twist parameter and may be the more convenient setting for statistical questions about CM curves, such as average ranks or Selmer groups, because the counting constants $C(j)$ are explicit in that family.","The dominance of the order $\\mathbb{Z}[( -1+\\sqrt{-3})/2]$ among the thirteen class-number-one orders suggests a general phenomenon: among CM structures over $\\mathbb{Q}$, the order with the largest unit group dominates at the level of natural density; one could test whether analogous dominance holds for CM abelian varieties of higher dimension."],"forward_implications":["The natural density of CM elliptic curves in the full family $\\mathcal{E}$ is $0$, so complex multiplication is asymptotically negligible among minimal short Weierstrass models ordered by naive height.","Inside the CM family, the curves with $j=0$ have natural density $1$, so asymptotically every CM elliptic curve over $\\mathbb{Q}$ has endomorphism ring $\\mathbb{Z}[( -1+\\sqrt{-3})/2]$.","The number of CM curves up to height $X$ grows like $\\frac{2}{3^{3/2}\\zeta(6)}X^{1/2}$, much slower than the $X^{5/6}$ growth of the total family.","In the twist-based family $\\mathcal{E}^{\\mathrm{cm}}_T$, the $j=0$ curves again have density $1$, and the counts for every $j\\in J^{\\mathrm{cm}}$ have explicit main terms $C(j)X^{1/m(j)}$ with $C(j)=\\frac{2}{\\zeta(n(j))}h_{\\mathrm{naive}}(E_j)^{-1/m(j)}$."],"supporting_citations":[{"why":"Supplies the total-family count #E(X) that sets the growth baseline against which the CM density zero is measured.","marker":"[Bru92]"},{"why":"Provides the k-free integer counting theorem used to get the main terms for j=0 and j=1728.","marker":"[MV07]"},{"why":"Classifies the thirteen CM orders of class number one and lists their j-invariants, the set J_cm used in the decomposition.","marker":"[Cox22]"},{"why":"Gives the degree/class-number formula that reduces CM over Q to class number one and fixes the possible j-invariants.","marker":"[Sil94]"},{"why":"Supplies the twist classification used to construct the alternative representative family E^T.","marker":"[Sil09]"},{"why":"One of the independent solutions of the class-number-one problem that determines the nine imaginary quadratic fields involved.","marker":"[Hee52]"},{"why":"Completes the proof that there are exactly nine imaginary quadratic fields of class number one, fixing the list of CM orders.","marker":"[Sta67b, Sta67a]"}],"fun_headline_variants":["CM curves density zero; j=0 takes all mass","CM elliptic curves: rare, and almost all j=0","Asymptotically, CM curves vanish; j=0 remains","Zero density CM; j=0 captures every limit point","All CM curves become j=0 in the limit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing estimate is Lemma 2.3's bound that a cuspidal cubic $y^2=a x^3$ with fixed rational $a$ has only $O(T^{1/2})$ integral points with $|x|\\le T$; if this bound failed, the eleven CM $j$-invariants other than $0$ and $1728$ could contribute a larger share and the claim that the $j=0$ curves form 100% of the CM family asymptotically could fail.","fun_headline_variants_meta":{"raw":{"variants":["CM curves density zero; j=0 takes all mass","CM elliptic curves: rare, and almost all j=0","Asymptotically, CM curves vanish; j=0 remains","Zero density CM; j=0 captures every limit point","All CM curves become j=0 in the limit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000371,"raw_usage":{"total_tokens":2052,"prompt_tokens":1080,"completion_tokens":972,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":696,"completion_tokens_details":{"reasoning_tokens":890}},"tokens_in":696,"tokens_out":972,"duration_ms":9995,"temperature":1.0,"reasoning_tokens":890,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:20:01.331426+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $\\#\\mathcal{E}_j(X)$ for one fixed CM $j$-invariant $j\\neq 0,1728$, such as $j=-3375$, up to $X=10^{12}$ using the counting code described in the paper; if the count grows faster than a constant multiple of $X^{1/6}$ (for instance like $X^{1/3}$ or $X^{1/2}$), the pivotal Lemma 2.3 bound fails and the 100% $j=0$ density conclusion would not follow. A more direct check is to search for integral points on $y^2=\\frac{1728-4j}{27j}x^3$ with $|x|\\le T$ and confirm whether their number exceeds the bound $2\\sigma_0(q)\\sqrt{|pq|}\\,T^{1/2}$ from the lemma.","supporting_citations":[],"review_version":1}