Home Panchromatic colorings of random hypergraphs
Article
Licensed
Unlicensed Requires Authentication

Panchromatic colorings of random hypergraphs

  • D. A. Kravtsov EMAIL logo , N. E. Krokhmal and D. A. Shabanov
Published/Copyright: February 16, 2021

Abstract

We study the threshold probability for the existence of a panchromatic coloring with r colors for a random k-homogeneous hypergraph in the binomial model H(n, k, p), that is, a coloring such that each edge of the hypergraph contains the vertices of all r colors. It is shown that this threshold probability for fixed r < k and increasing n corresponds to the sparse case, i. e. the case p=cn/(nk) for positive fixed c. Estimates of the form c1(r, k) < c < c2(r, k) for the parameter c are found such that the difference c2(r, k) − c1(r, k) converges exponentially fast to zero if r is fixed and k tends to infinity.


Originally published in Diskretnaya Matematika (2019) 31, №2, 84–113 (in Russian).


Funding statement: The study was performed under the RFBR grant No. 16-11-10014.

References

[1] M. Dyer, A. Frieze, C. Greenhill, “On the chromatic number of a random hypergraph”, J. Comb. Theory, Ser. B, 113 (2015), 68–122.10.1016/j.jctb.2015.01.002Search in Google Scholar

[2] N. Alon, J. Spencer, “A note on coloring random k-sets”, Unpublished manuscript, http://www.cs.tau.ac.il/~nogaa/PDFS/kset2.pdf.Search in Google Scholar

[3] D. Achlioptas, J.H. Kim, M. Krivelevich, P. Tetali, “Two-colorings random hypergraphs”, Random Struct. Algor., 20:2 (2002), 249–259.10.1002/rsa.997Search in Google Scholar

[4] D. Achlioptas, C. Moore, “Random k-SAT: two moments suffice to cross a sharp threshold”, SIAM J. Comput., 36:3 (2005), 740–762.10.1137/S0097539703434231Search in Google Scholar

[5] A. Coja-Oghlan, L. Zdeborová, “The condensation transition in random hypergraph 2-coloring”, Proc. 23rd Annu. ACM–SIAM Symp. Discr. Algor., SIAM, 2012, 241–250.10.1137/1.9781611973099.22Search in Google Scholar

[6] P. Erdős, L. Lovász, “Problems and results on 3-chromatic hypergraphs and some related questions”, Infinite and Finite Sets, Colloq. Math. Soc. J. Bolyai, 10, Amsterdam: North Holland, 1973, 609–627.Search in Google Scholar

[7] A.V. Kostochka, “On a theorem by Erdős, Rubin and Taylor on choosability of complete bipartite graphs”, Electr. J. Comb., 9 (2002).10.37236/1670Search in Google Scholar

[8] D.A. Shabanov, “On a generalization of Rubin’s theorem”, J. Graph Theory, 67:3 (2011), 226–234.10.1002/jgt.20527Search in Google Scholar

[9] D.A. Kravtsov, N.E. Krokhmal, D.A. Shabanov, “Panchromatic 3-colorings of random hypergraphs”, Eur. J. Comb., 78 (2019), 28–43.10.1016/j.ejc.2019.01.006Search in Google Scholar

[10] D.A. Shabanov, “The existence of panchromatic colourings for uniform hypergraphs”, Sb. Math., 201:4 (2010), 607–630.10.1070/SM2010v201n04ABEH004084Search in Google Scholar

[11] D.A. Shabanov, “On the concentration of the chromatic number of a random hypergraph”, Doklady Mathematics, 96:1 (2017), 321–325.10.1134/S1064562417040044Search in Google Scholar

[12] H. Hatami, M. Molloy, “Sharp thresholds for constraint satisfaction problems and homomorphisms”, Random Struct. Algor., 33:3 (2008), 310–332.10.1002/rsa.20225Search in Google Scholar

[13] D. Cherkashin, “Note on panchromatic colorings”, Discrete Mathematics, 341:13 (2018), 652–657.10.1016/j.disc.2017.10.030Search in Google Scholar

4.6

4.6 Appendix

In the Appendix we prove the auxiliary technical statements.

Assertion 2

There existsk0 < such that for allk > k0 and 3r<0.1kwe have

k24r(r2(4/3)r+2/3r2)k2<(r1r)k.

Proof

Note that this inequality is valid for all r ⩽ 11 and sufficiently large k. For r > 11 we have r2 − (4/3)r + 2/3 < r2 − (5/4)r and

k24r(r2(4/3)r+2/3r2)k2<k24r(r5/4r)k2k24r(r5/4r)k(65)2(65)2k24re5k4r=(65)2k2exp{k4r+(ln4)r+kr(r1)}ekr1.

If r > 11, then kr(r1)<k8r, and for r<0.1k we have k8r>(5/4)k. Hence,

k24r(r2(4/3)r+2/3r2)k2(65)2k2exp{(5/4)k+(ln4)r}ekr1<(65)2k2exp{k}ekr1<ekr1<(r1r)k

for all sufficiently large k.

Assertion 3

There existsk0 < such that for allk > k0 and 3r<0.1kthe following inequalities are valid:

2k2lnrr(r22r+2r(r1))k(r2r22r+2)2r218,2rlnr(r2r)k1162r2.

Proof

In the first inequality the factors of the left hand side of (2lnr/r)(r2r22r+2)2 are bounded, so it is sufficient to show that the product of two other tends to zero uniformy over all r<0.1k as k tends to infinity. The relation

k2(r22r+2r(r1))k=k2(1r2r(r1))kk2exp{(r2)kr(r1)}k2e10k+o(k)0fork

follows from the monotone descreasing of (r − 2)/(r2r) with respect to r.

It is sufficient to show that for large k the value 2r3lnr(r2r)k is small for all sufficiently large k:

2r3lnr(r2r)k2k4exp{2kr}2k4e20k0fork.

Assertion 4

There existsk0 < such that for all k > k0 and 3r<0.1kthe following inequality is valid:

4lnrrk22(r2rrk1/8r(r1))k(r2r2rrk1/8)2120r2.

Proof

It is sufficient to show that k4(r2rrk1/8r(r1))k tends to zero uniformly in r if k tends to infinity. The estimates

k4(r2rrk1/8r(r1))k=k4(1rk1/8r(r1))kk4exp{rk7/8r(r1)}k4e103/2k8+o(k8)0fork

follows from the fact that the function r/(r2r) is decreasing in r.

Assertion 5

There existsk0 < ∞ such that for allk > k0, 3r<0.1k and arbitrary numbers δij ∈ [0, r−3/2k−1/8] we have

(er1/2+ji(r2r1+rr1δij)k)(1r(r1r)k)2112r.

Proof

Note that er1/2<(1+r1/2)1=11r+1. Further,

r(r1r)krek/re10k+O(lnk);ji(r2r1+rr1δij)kjiekr1+rr1δijkrekr1+kr(r1)r3/2k1/8e10k(1+o(1)).

So, the estimated value does not exceed

(11r+1+e10k+O(lnk))(1+e10k(1+o(1))),

this expression for all sufficiently large k implies the desired upper bound 11/(2r).

Assertion 6

There existsk0 < ∞ such that for allk > k0, 3r<0.1kthe following inequality is valid:

2(r1)rlnr(r2r)k+0.01lnrrk(rr+1)klnrr(rr+1)k.

Proof

From the conditions on the parameters we deduce that

2(r1)rlnr(r2r)k+0.01lnrrk(rr+1)k0.2klnr(r2r)k+0.01lnrrk(rr+1)k0.2klnr(rr+1)2k+0.01lnrrk(rr+1)k.

Further, r(rr+1)krek/(r+1) tends to zero as k tends to infinity, hence it is smaller than 1/40 for large k. Therefore

2(r1)rlnr(r2r)k+0.01lnrrk(rr+1)k0.005klnrr(rr+1)k+0.01lnrrk(rr+1)k=0.005lnrrk(rr+1)klnrr(rr+1)k,

where the last inequality also is satisfied for all sufficiently large k.

Assertion 7

There existsk0 < ∞ such that for allk > k0, 3r<0.1kwe have

(r1)2rrkr13(r1)lnr(r1r)k7rlnrr1k2(rr+1)2k+0.01lnrrk(rr+1)k0.01lnr4rk(rr+1)k.

Proof

It is sufficient to show that each of the first three negative summands is 4 times smaller than the last summand. Let us compare each of them with the last summand. Begin with the first one:

(r1)2rrkr1(0.01lnrrk(rr+1)k)1=100(r1)2klnrrkr1(r+1r)k100(r1)2klnrexp{klnrr1+kr}=100(r1)2klnrexp{krlnrr+1r(r1)}100klnkexp{5klnk(1+o(1))}14

for all sufficiently large k. Now consider the second summand:

3(r1)lnr(r1r)k(0.01lnrrk(rr+1)k)1=300(r1)rk(11r2)k300k100kek/r23e100<14.

It remains to consider the third summand:

7rlnrr1k2(rr+1)2k(0.01lnrrk(rr+1)k)1=700r2kr1(rr+1)k7k3ek/(r+1)7k3e10k(1+o(1))<14

for all sufficiently large k.

Assertion 8

There existsk0 < ∞ such that for allk > k0, 3r<0.1kand arbitraryδii[0,(rr+1)k], i = 1, …, r, the following inequality is valid:

|i=1r[δiilnδii+ln(rr1)δiiδii]|×(kr1i=1rδii+r2(r1r)k+14k2r(r2(r+1)2)k)3k2r(r2(r+1)2)k.

Proof

Note that lnδii<kr+1 is sufficiently large in the absolute value, so for large k we may suppose that |δiilnδii+ln(rr1)δiiδii|2|δiilnδii|2kln(r+1r)(rr+1)k2kr+1(rr+1)k. Consequently, the left hand side of our inequality does not exceed

2krr+1(rr+1)k(krr1(rr+1)k+r2(r1r)k+14k2r(r2(r+1)2)k).

In view of condition on the relation between r and k the maximal summand in the second parenthesis is the first one. Hence, the left hand side of our inequality does not exceed

2krr+1(rr+1)k3krr1(rr+1)k3k2r(r2(r+1)2)k.

The assertion 8 is proved.

Received: 2018-02-11
Published Online: 2021-02-16
Published in Print: 2021-02-23

© 2021 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 29.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/dma-2021-0003/html?lang=en
Scroll to top button