Skip to main content
Log in

Nonexistence of nontrivial tight 8-designs

  • Published:
Journal of Algebraic Combinatorics Aims and scope Submit manuscript

Abstract

Tight t-designs are t-designs whose sizes achieve the Fisher type lower bound. We give a new necessary condition for the existence of nontrivial tight designs and then use it to show that there do not exist nontrivial tight 8-designs.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1

Similar content being viewed by others

References

  1. Bannai, E.: On tight designs. Q. J. Math. 28(4), 433–448 (1977). doi:10.1093/qmath/28.4.433

    Article  MathSciNet  MATH  Google Scholar 

  2. Bannai, E., and Ito, T.: Unpublished (1977)

  3. Bremner, A.: A diophantine equation arising from tight \(4\)-designs. Osaka J. Math. 16(2), 353–356 (1979). http://hdl.handle.net/11094/8744

  4. Delsarte, P.: An algebraic approach to the association schemes of coding theory. Philips Res. Repts. Suppl. 10 (1973)

  5. Dukes, P., Short-Gershman, J.: Nonexistence results for tight block designs. J. Algebr. Comb. 38(1), 103–119 (2013). doi:10.1007/s10801-012-0395-8

    Article  MathSciNet  MATH  Google Scholar 

  6. Edgar, T., Spivey, M.Z.: Multiplicative functions generalized binomial coefficients, and generalized catalan numbers. J. Integer Seq. 19(2), 737–744 (2016)

    MathSciNet  MATH  Google Scholar 

  7. Hall Jr., M.: Combinatorial Theory, 2nd edn. Wiley, New York (1998)

    MATH  Google Scholar 

  8. Hilliker, D.L., Straus, E.G.: Determination of bounds for the solutions to those binary Diophantine equations that satisfy the hypotheses of Runge’s theorem. Trans. Am. Math. Soc. 280(2), 637–657 (1983). doi:10.1090/S0002-9947-1983-0716842-3

    MathSciNet  MATH  Google Scholar 

  9. Hindry, M., and Silverman, J.H.: Diophantine Geometry: An Introduction (2000). http://www.springer.com/us/book/9780387989754

  10. Lorenzini, D., and Xiang, Z.: Integral points on variable separated curves (2015)

  11. Peterson, C.: On tight \(6\)-designs. Osaka J. Math. 14(2), 417–435 (1977). http://hdl.handle.net/11094/3990

  12. Ray-Chaudhuri, D.K., and Wilson, R.M.: On \(t\)-designs. Osaka J. Math. 12(3), 737–744 (1975). http://hdl.handle.net/11094/7296

  13. Runge, C.: Über ganzzahlige Lösungen von Gleichungen zwischen zwei Veränderlichen. Journal für die reine und angewandte Mathematik, 100, 425–435 (1887). http://eudml.org/doc/148675

  14. Stroeker, R.J.: On the diophantine equation \((2y^2-3)^2 = x^2(3x^2-2)\) in connection with the existence of non-trivial tight \(4\)-designs. Indag. Math. 84(3), 353–358 (1981)

    Article  MathSciNet  MATH  Google Scholar 

  15. Walsh, P.G.: A quantitative version of Runge’s theorem on Diophantine equations. Acta Arith. 62(2), 157–172 (1992). http://eudml.org/doc/206487

  16. Zannier, U.: Lecture Notes on Diophantine Analysis (2015)

Download references

Acknowledgements

It is my great pleasure to thank Eiichi Bannai for introducing me to the subject and suggesting this problem to me. I also thank Etsuko Bannai, Dino Lorenzini, Yaokun Wu and Jiacheng Xia for useful comments and discussions. I gratefully acknowledge financial support from the Office of the Vice President for Research at the University of Georgia, and from the Research and Training Group in Algebraic Geometry, Algebra and Number Theory.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ziqing Xiang.

Appendices

A The thirteen degree polynomial

\(f_4(v, k) := -16384 k^{12} v+65536 k^{12}+98304 k^{11} v^2-393216 k^{11} v-253952 k^{10} v^3+786432 k^{10} v^2+1744896 k^{10} v-3309568 k^{10}+368640 k^9 v^4-327680 k^9 v^3-8724480 k^9 v^2+16547840 k^9 v-328320 k^8 v^5-1102464 k^8 v^4+17194752 k^8 v^3-21567744 k^8 v^2-49810560 k^8 v+62323584 k^8+182784 k^7 v^6+2050560 k^7 v^5-16432128 k^7 v^4-13016064 k^7 v^3+199242240 k^7 v^2-249294336 k^7 v-61184 k^6 v^7-1642240 k^6 v^6+6536960 k^6 v^5+58253568 k^6 v^4-293538048 k^6 v^3+209662720 k^6 v^2+511604992 k^6 v-488998144 k^6+10752 k^5 v^8+698880 k^5 v^7+1258752 k^5 v^6-59703552 k^5 v^5+183266304 k^5 v^4+243542016 k^5 v^3-1534814976 k^5 v^2+1466994432 k^5 v-640 k^4 v^9-143664 k^4 v^8-2296192 k^4 v^7+27050224 k^4 v^6-7038496 k^4 v^5-582955856 k^4 v^4+1856597696 k^4 v^3-1428764528 k^4 v^2-1015706784 k^4 v+974873344 k^4+7520 k^3 v^9+772608 k^3 v^8-2875616 k^3 v^7-58917568 k^3 v^6+469164960 k^3 v^5-1155170432 k^3 v^4+412538336 k^3 v^3+2031413568 k^3 v^2-1949746688 k^3 v+336 k^2 v^{10}-52816 k^2 v^9-1582560 k^2 v^8+27560816 k^2 v^7-127930016 k^2 v^6+28759472 k^2 v^5+1497511456 k^2 v^4-4944873072 k^2 v^3+6922441360 k^2 v^2-4733985888 k^2 v+1506333312 k^2-2352 k v^{10}+203472 k v^9-764688 k v^8-24513072 k v^7+293023248 k v^6-1459281552 k v^5+3929166288 k v^4-5947568016 k v^3+4733985888 k v^2-1506333312 k v+45 v^{11}+972 v^{10}-191952 v^9+2961396 v^8-14780538 v^7-18769932 v^6+544096980 v^5-2755473732 v^4+7281931941 v^3-11097146016 v^2+9310949028 v-3408102864\).

B Proof of Proposition 5.3

Lemmas B.1, B.2, B.3, and B.4, prove Proposition 5.3. The proofs of these lemmas are not difficult once the appropriate auxiliary polynomials g has been explicitly identified. Such auxiliary polynomials were obtained through an ad-hoc procedure that we will not describe here.

Lemma B.1

If \(v \in [2k, 9k]\) and \(k \ge 20000\), then \(f_4(v, k) > 0\).

Proof

Let

$$\begin{aligned} g(v, k):= & {} 65535 k^{12} - 16404 k^{12} v + 98304 k^{11} v^2 - 253969 k^{10} v^3 \\&+\, 368584 k^9 v^4 - 328320 k^8 v^5 + 182701 k^7 v^6 - 61184 k^6 v^7 \\&+\,10744 k^5 v^8 - 640 k^4 v^9 + 335 k^2 v^{10} + 45 v^{11}. \end{aligned}$$

View \(f_4(v, k)\) and g(vk) as polynomials in v with coefficients in \({{\mathrm{\mathbb {R}}}}[k]\). Using the fact that \(k \ge 20000\), we can check that for every \(i \ge 0\), the coefficient of \(v^i\) in \(f_4(v, k)\) is no smaller than that in g(vk), and for some i, the coefficient of \(v^i\) in \(f_4(v, k)\) is strictly larger than that in g(vk). So, we find that \(f_4(v, k) > g(v, k)\) since \(v \ge 2 k > 0\).

Let \(t := v / k\), so that \(t \in [2, 9]\). We have

$$\begin{aligned} f_4(v, k) > g(v, k)= & {} g(t k, k) \\= & {} h_{13}(t) k^{13} + h_{12}(t) k^{12} + h_{11}(t) k^{11}, \end{aligned}$$

where

$$\begin{aligned} h_{13}(t)&:= -16404 t + 98304 t^2 - 253969 t^3 + 368584 t^4 \\&\quad - 328320 t^5 + 182701 t^6 - 61184 t^7 + 10744 t^8 - 640 t^9, \\ h_{12}(t)&:= 65535 + 335 t^{10}, \\ h_{11}(t)&:= 45 t^{11}. \end{aligned}$$

One verifies that \(h_{13}\), \(h_{12}\) and \(h_{11}\) are positive when \(t \in [2, 9]\), and then the result follows. \(\square \)

Recall the definitions of real numbers a and b in Eq. (5.1):

$$\begin{aligned} {\left\{ \begin{array}{ll} a = \frac{2}{1 - \root 4 \of {\frac{3}{8}}} = \frac{2}{5} (8 + 2 \sqrt{6} + \sqrt{48 + 22 \sqrt{6}}) \approx 9.1971905725, \\ b = \frac{23}{500} \left( 249 + 86 \sqrt{6} + \sqrt{171312 + 70918 \sqrt{6}} \right) \approx 48.1640392521. \\ \end{array}\right. } \end{aligned}$$

Lemma B.2

If \(v \in [9 k, a k + b]\) and \(k \ge 100\), then \(f_4(v, k) > 0\).

Proof

Let \(t := a k + b - v\), so that \(t \in [0, (a - 9) k + b] \subseteq [0, 0.9 k]\) when \(k \ge 100\). Let

$$\begin{aligned} g(t, k)&:= 13700000000 t k^{12} - 13400000000 t^2 k^{11} - 140000000 t^3 k^{10} \\&\quad - 1380000000 t^4 k^9 - 6000000 t^5 k^8 - 21000000 t^6 k^7 \\&\quad - 10000 t^7 k^6 - 73000 t^8 k^5 - 40000 t^9 k^3 - 7 t^9 k^4 \\&\quad - 50000 t^{10} k - 200 t^{10} k^2 - 312000000000 t^{11}. \end{aligned}$$

View \(f_4(a k + b - t, k)\) and g(tk) as polynomials in k with coefficients in \({{\mathrm{\mathbb {R}}}}[t]\). Using the fact that \(t \ge 0\), we can check that for every \(i \ge 0\), the coefficient of \(k^i\) in \(f_4(a k + b - t, k)\) is no smaller than that in g(tk), and for some i, the coefficient of \(k^i\) in \(f_4(a k + b - t, k)\) is strictly larger than that in g(tk). So, we find that \(f_4(a k + b - t, k) > g(t, k)\) since \(k > 0\).

Let \(s := t / k\), so that \(s \in [0, 0.9]\). We have

$$\begin{aligned} f_4(v, k)= & {} f_4(a k + b - t) > g(t, k) = g(s k, k) \\= & {} h_{13}(s) k^{13} - h_{12}(s) k^{12} - h_{11}(s) k^{11}, \end{aligned}$$

where

$$\begin{aligned} h_{13}(s)&:= 13700000000 s - 13400000000 s^2 - 140000000 s^3 \\&\quad - 1380000000 s^4 - 6000000 s^5 - 21000000 s^6 - 10000 s^7 \\&\quad - 73000 s^8 - 7 s^9, \\ h_{12}(s)&:= 40000 s^9 + 200 s^{10}, \\ h_{11}(s)&:= 50000 s^{10} + 312000000000 s^{11}. \end{aligned}$$

One verifies that \(h_{13}\), \(h_{12}\) and \(h_{11}\) are positive when \(s \in [0, 0.9]\). Since \(k \ge 100\)

$$\begin{aligned} f_4(v, k)&> h_{13}(s) k^{13} - h_{12}(s) k^{12} - h_{11}(s) k^{11} \\&\ge (h_{13}(s) - h_{12}(s) / 100 - h_{11}(s) / 100^2) k^{13} \\&= \bigg (13700000000 s - 13400000000 s^2 - 140000000 s^3 \\&\quad - 1380000000 s^4 - 6000000 s^5 - 21000000 s^6 - 10000 s^7 \\&\quad - 73000 s^8 - 407 s^9 - 7 s^{10} - 31200000 s^{11}\bigg ) k^{13}. \end{aligned}$$

The result follows from the fact that the right-hand side of the expression above is positive when \(s \in [0, 0.9]\).

Lemma B.3

If \(a k + b + \frac{1}{100} \le v \le 10 k\) and \(k \ge 10^5\), then \(f_4(v, k) < 0\).

Proof

Let \(t := v - a k - b\), so that \(t \in [\frac{1}{100}, (10 - a) k - b] \subseteq [\frac{1}{100}, k]\) when \(k \ge 10^5\). Let

$$\begin{aligned} g(t, k)&:= -137892000 k^{12} + 13318642886180 k^{11} + 713748202323829 k^{10} \\&\quad + 48837673261525668 k^9 + 5098485316801241991 k^8 \\&\quad + 980132640412645508268 k^7 + 488910709935302976594934 k^6 \\&\quad + 1305009906289977795675277621 k^5 \\&\quad + 151418572251274917743210971453210 k^4 \\&\quad + 64000 t^9 k^3 + 2000 t^{10} k^2 + 7000 t^{10} k + 127 t^{11}. \end{aligned}$$

View \(f_4(a k + b + t, k)\) and g(tk) as polynomials in k with coefficients in \({{\mathrm{\mathbb {R}}}}[t]\). Using the fact that \(t \ge \frac{1}{100}\), we can check that for every \(i \ge 0\), the coefficient of \(k^i\) in \(f_4(a k + b + t, k)\) is no larger than that in g(tk), and for some i, the coefficient of \(k^i\) in \(f_4(a k + b + t, k)\) is strictly smaller than that in g(tk). So, we find that \(f_4(a k + b + t, k) < g(t, k)\) since \(k > 0\). It follows from \(t \le k\) that \(g(t, k) \le g(k, k)\). It is easy to show the one variable polynomial g(kk) takes negative value when \(k \ge 10^5\). Thus, \(f_4(v, k)< g(t, k) \le g(k, k) < 0\). \(\square \)

Lemma B.4

If \(v \in [9.24 k, 0.8 k^2]\) and \(k \ge 10^5\), then \(f_4(v, k) < 0\).

Proof

Let

$$\begin{aligned} g(v, k)&:= 65536 k^{12} - 16384 k^{12} v + 98312 k^{11} v^2 - 253952 k^{10} v^3 \\&\quad + 368640 k^9 v^4 - 328299 k^8 v^5 + 182784 k^7 v^6 - 61177 k^6 v^7 \\&\quad + 10752 k^5 v^8 - 639 k^4 v^9 + 336 k^2 v^{10} + 45 v^{11}. \end{aligned}$$

View \(f_4(v, k)\) and g(vk) as polynomials in v with coefficients in \({{\mathrm{\mathbb {R}}}}[k]\). Using the fact that \(k \ge 10^5\), we can check that for every \(i \ge 0\), the coefficient of \(v^i\) in \(f_4(v, k)\) is no larger than that in g(vk), and for some i, the coefficient of \(v^i\) in \(f_4(v, k)\) is strictly smaller than that in g(vk). So, we find that \(f_4(v, k) < g(v, k)\) since \(v > 0\).

Let \(t := v / k\), so that \(t \in [9.24, 0.8k]\). Let

$$\begin{aligned} h(x)&:= - 1 + 16384 x - 98312 x^2 + 253952 x^3 - 368640 x^4 \\&\quad + 328299 x^5 - 182784 x^6 + 61177 x^7 - 10752 x^8 + 639 x^9. \end{aligned}$$

We have

$$\begin{aligned}&f_4(v, k) < g(v, k) = g(t k, k) \nonumber \\ =&- \big (k - 65536\big ) k^{12} - \big (h(t) - 298 t^9\big ) k^{13} \nonumber \\&- \big (298 - 336 (t / k) - 45 (t / k)^2\big ) t^9 k^{13} \end{aligned}$$
(B.1)
$$\begin{aligned} =&- \big (k - 65536\big ) k^{12} - \big (h(t) - 0.1 t^9\big ) k^{13} \nonumber \\&- \big (0.1 - 336 (t / k) - 45 (t / k)^2\big ) t^9 k^{13}. \end{aligned}$$
(B.2)

It suffices to prove that the coefficients of \(k^{12}\), \(k^{13}\) and \(t^9 k^{13}\) in Eq. (B.1) are all negative, or those coefficients in Eq. (B.2) are all negative. Since \(k \ge 10^5\), we have \(0.00026 k^2 \ge 26 k\). So, either \(v \ge 26 k\) or \(v \le 0.00026 k^2\).

Case 1: \(v \in [26 k, 0.8 k^2]\), so that \(t \in [26, 0.8 k]\) and \(t / k \in [0, 0.8]\).

Consider Eq. (B.1). The result follows from the following facts:

  1. (i)

    \(k - 65536 \ge 0\);

  2. (ii)

    \(h(t) - 298 t^9 \ge 0\) when \(t \ge 26\);

  3. (iii)

    \(298 - 336 (t / k) - 45 (t / k)^2 \ge 0\) when \(t / k \in [0, 0.8]\).

Case 2: \(v \in [9.24k, 0.00026 k^2]\) so that \(t \in [9.24, 0.00026 k]\) and \(t / k \in [0, 0.00026]\).

Consider Eq. (B.2). The result follows from the following facts:

  1. (i)

    \(k - 65536 \ge 0\);

  2. (ii)

    \(h(t) - 0.1 t^9 \ge 0\) when \(t \ge 9.24\);

  3. (iii)

    \(0.1 - 336 (t / k) - 45 (t / k)^2 \ge 0\) when \(t / k \in [0, 0.00026]\). \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Xiang, Z. Nonexistence of nontrivial tight 8-designs. J Algebr Comb 47, 301–318 (2018). https://doi.org/10.1007/s10801-017-0776-0

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10801-017-0776-0

Keywords

Navigation