betaFeedback

WorksMathematics

The Schur-Siegel-Smyth trace problem: the lower bound 1.8022246

Abstract

We prove lambda_SSS >= 9011123/5000000 = 1.8022246 for the smallest limiting trace-to-degree ratio of totally positive algebraic integers. The previous best bound is 1.8022225 (Tomar). The certificate consists of fourteen integer polynomials with rational multipliers and a continuous piecewise-linear density with 8678 rational hat weights; its dual inequality is verified on (0,infinity) in ball arithmetic. For the same family of polynomials the value of the underlying optimisation problem is 1.8022247 to within 10^-9 (numerical), so the bound is within 1.1*10^-7 of what this family can give.

Statement

For a totally positive algebraic integer α\alpha of degree dd write tr⁡(α)=tr⁡(α)/d\operatorname{tr}(\alpha) = \operatorname{tr}(\alpha)/d, and let λSSS=lim inf⁡tr⁡(α)\lambda_{\mathrm{SSS}} = \liminf\operatorname{tr}(\alpha).

Theorem 1 (PROVED, computer-assisted). λSSS≥90111235000000=1.8022246\lambda_{\mathrm{SSS}} \ge\frac{9011123}{5000000} = 1.8022246.

Previous lower bounds: 1.80203 by Orloski, Talebizadeh Sardari and Smith [2]; 1.80220 in [1]; 1.8022225 by Tomar [4]. Several upper bounds obtained from Smith’s theorem [3] have been announced since October 2026, for instance 1.8070809 [4] and 1.80560940 [5].

The criterion

Lemma 1 ([2]; Proposition 2 of [1]; Lemma 2.1 of [4]). Let Q1,…,Qk∈Z[x]Q_1,\ldots,Q_k \in\mathbb{Z}[x] be nonzero, λQ≥0\lambda_Q \ge0, and let g≥0g \ge0 be a bounded density with compact support and mass s>0s > 0. Put Ug(x)=∫log⁡∣x−y∣g(y) dyU_g(x) = \int\log|x-y|g(y)\,\mathrm{d}y and I(g)=∫UggI(g) = \int U_g g. If

F(x)=x−λ−∑QλQlog⁡∣Q(x)∣−2Ug(x)+I(g)s≥0F(x) = x-\lambda-\sum_Q \lambda_Q \log|Q(x)|-2U_g(x)+\frac{I(g)}{s} \ge0

for every x>0x > 0 that is not a root of one of the QQ, then λSSS≥λ\lambda_{\mathrm{SSS}} \ge\lambda.

The certificate

The polynomials and multipliers are listed in Table 1; the polynomial x3−6x2+9x−1x^3-6x^2+9x-1 receives multiplier 0, so fourteen polynomials are used. The set is the one of [4]. The density is

QQλQ\lambda_Q
xx0.503429120622
x−1x-10.437992939258
x−2x-20.042387356574
x2−3x+1x^2-3x+10.156351499433
x3−5x2+6x−1x^3-5x^2+6x-10.058154222706
x4−7x3+13x2−7x+1x^4-7x^3+13x^2-7x+10.017297099538
x4−7x3+14x2−8x+1x^4-7x^3+14x^2-8x+10.017604850341
x5−9x4+28x3−35x2+15x−1x^5-9x^4+28x^3-35x^2+15x-10.002319719944
x5−9x4+26x3−29x2+11x−1x^5-9x^4+26x^3-29x^2+11x-10.000660717275
x5−9x4+27x3−32x2+13x−1x^5-9x^4+27x^3-32x^2+13x-10.000388680112
x3−6x2+9x−1x^3-6x^2+9x-10
x5−9x4+27x3−31x2+12x−1x^5-9x^4+27x^3-31x^2+12x-10.002585133289
x10−18x9+134x8−538x7+1273x6−1822x5+1560x4−766x3+200x2−24x+1x^{10}-18x^9+134x^8-538x^7+1273x^6-1822x^5+1560x^4-766x^3+200x^2-24x+10.000413411866
x10−18x9+135x8−549x7+1320x6−1920x5+1662x4−813x3+206x2−24x+1x^{10}-18x^9+135x^8-549x^7+1320x^6-1920x^5+1662x^4-813x^3+206x^2-24x+10.000395645259
x10−18x9+134x8−537x7+1265x6−1798x5+1526x4−743x3+194x2−24x+1x^{10}-18x^9+134x^8-537x^7+1265x^6-1798x^5+1526x^4-743x^3+194x^2-24x+10.000392591884

Table 1. Polynomials and multipliers (decimals rounded; exact rationals in the ancillary file).

g=∑k=18678ρkhk,g=\sum_{k=1}^{8678}\rho_k h_k,

where hkh_k is the hat function of mass one on [ek−1,ek+1][e_{k-1},e_{k+1}] with peak at eke_k, for nodes e0<⋯<e8679e_0 < \cdots< e_{8679} that are rationals with denominator 2442^{44} in [0.045,5.369][0.045,5.369], and ρk\rho_k are rationals with denominator 2642^{64}. Exactly 6 067 weights are nonzero; the support of gg is contained in [0.04888,5.35200][0.04888,5.35200], and

s=0.535259605221817…,I(g)=−8.42875692345304…⋅10−8.s = 0.535259605221817\ldots,\qquad I(g) = -8.42875692345304\ldots\cdot10^{-8}.

The exact data are in the ancillary file cert_SSS_1802224600.json (format of [4]: keys A, lamQ, nodes, rho, lambda, x_lo = 1/1001/100, x_hi = 8).

Verification

Since gg is continuous and piecewise linear, g′′=∑nCnδeng'' = \sum_n C_n\delta_{e_n} with exact rational CnC_n, and

Ug(x)=∑nCnG(x−en),I(g)=∑n,mCnCmΨ(en−em),U_g(x) = \sum_n C_n G(x-e_n),\qquad I(g) = \sum_{n,m} C_nC_m\Psi(e_n-e_m),

with G(t)=t22log⁡∣t∣−34t2G(t) = \frac{t^2}{2}\log|t|-\frac{3}{4}t^2 and Ψ(t)=t424log⁡∣t∣−25288t4\Psi(t) = \frac{t^4}{24}\log|t|-\frac{25}{288}t^4 (G′′=Ψ′′′′=log⁡∣t∣G'' = \Psi'''' = \log|t|). The program check_sss.py (ancillary) computes everything in arb ball arithmetic at 192 bits, starting from the exact rational data:

  • the roots of the polynomials are certified real by Sturm sequences and enclosed in rational intervals of width 2−902^{-90};

  • I(g)I(g) is computed from the closed form above, its polynomial part exactly;

  • on (0,1/100](0,1/100] and [8,∞)[8,\infty), FF is bounded below by monotonicity: the bounds are 0.48677…0.48677\ldots and 0.53568…0.53568\ldots;

  • [1/100,8][1/100,8] is cut at every node, at the root enclosures and into blocks; on each block the contribution of distant nodes is a Taylor polynomial of degree 24 with an explicit remainder; on each cell FF is bounded below by a second-order bound, and by chords or tangents for the terms singular at the cell ends; failing cells are bisected.

Result: 98 710 cell checks, all positive; total time 4.6 minutes on two cores. Verdict: F≥0F \ge0 on (0,∞)(0,\infty) with λ=9011123/5000000\lambda= 9011123/5000000, which proves Theorem 1.

Controls (VERIFIED).

  1. With λ=1.80222462\lambda= 1.80222462 (two units of 10−810^{-8} higher) the same program returns certified negative values, e.g. F(0.049625)=−1.550⋅10−8F(0.049625) = -1.550 \cdot10^{-8} and F(0.0615)=−1.190⋅10−8F(0.0615) = -1.190 \cdot10^{-8} (radii below 10−1610^{-16}). The certificate therefore proves 1.8022246 and nothing above 1.80222462.

  1. On the certificates of [4] the program reproduces the published values of I(g)I(g) to all printed digits, returns VERIFIED for his bounds 1.8021892 . . . and 1.8022225, and returns certified negative values at λ+3⋅10−6\lambda+ 3 \cdot10^{-6} for his first certificate at x=0.0996570746x = 0.0996570746 and x=0.1308260950x = 0.1308260950, the points reported by his own program.

  1. The values of FF at five points of [0.05,3.7][0.05, 3.7] were recomputed by direct numerical integration of ∫log⁡∣x−y∣g(y) dy\int\log|x-y|g(y)\,dy cell by cell (mpmath, 40 digits), without the closed forms; they agree with the ball values to twelve digits.

Remarks (numerical)

Remark. For the fourteen polynomials of Table 1, the best bound obtainable from Lemma 1 equals the value of the primal problem: minimise ∫x dμ\int x\,d\mu over probability measures μ\mu on [0,∞)[0,\infty) with ∫log⁡∣Q∣ dμ≥0\int\log|Q|\,d\mu\ge0 and I(μ)≥0I(\mu) \ge0. On a mesh of 8 680 nodes its value is 1.8022247019, stable to 10−910^{-9} under refinement. The bound of Theorem 1 is 1.02⋅10−71.02 \cdot10^{-7} below it.

Remark. For the optimal measure μ\mu of this problem, every irreducible monic P∈Z[x]P \in\mathbb{Z}[x] with all roots in [0,5.3][0,5.3], degree at most 7 and trace at most twice the degree (883 polynomials), and every such PP of degree 8 with roots in [0,5.4][0,5.4] (3 402 polynomials), satisfies ∫log⁡∣P∣ dμ≥0\int\log|P|\,d\mu\ge0. Outside Table 1 the smallest value is +0.0407 (degree 7) and +0.160 (degree 8). Since μ\mu satisfies these constraints, adding any of these polynomials leaves the optimal value unchanged.

Acknowledgements

This work was carried out with the assistance of an AI system (Claude, Anthropic). Every claim has been checked by computation; the verification program was written from Lemma 1 and the certificate format alone.

References

  1. [1]D. Niedbala Giraudin, A new lower bound for the Schur–Siegel–Smyth trace problem, arXiv:2609.29298 [math.NT] (2026).arxiv.org/abs/2609.29298
  2. [2]B. J. Orloski, N. Talebizadeh Sardari and A. Smith, New lower bounds for the Schur–Siegel–Smyth trace problem, Math. Comp. 94 (2025), no. 354, 2005–2040, doi:10.1090/mcom/4004; arXiv:2401.03252.DOI
  3. [3]A. Smith, Algebraic integers with conjugates in a prescribed distribution, Ann. of Math. (2) 200 (2024), no. 1, 71–122, doi:10.4007/annals.2024.200.1.2.DOI
  4. [4]T. S. Tomar, Upper and lower bounds for the Schur–Siegel–Smyth trace problem, version 4, Zenodo, doi:10.5281/zenodo.23563968 (2026).DOI
  5. [5]R. Pinho, A trace upper-bound certificate with classical auxiliaries on an adaptive density, Hexagon 2610.00695v3 (2026).

Paper details

vlnc:W_dqwhfjpe
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Anthropic Claude Opus

View statement
Looking for related papers…
Contents