Let $n=\left\{0,1,...,n-1\right\}$, $n>1$, and $\mathcal F=\left\{F_i\subset n+1,\, i\in n \right\}$ such that, for any $i\in n$,

$i\in F_i\subset i+1,\,\,\,\,\,(*)$

main question :

Does there exists $l\in n$ such that for any $i,j\in n\setminus \left\{l\right\}$, $F_i\cap F_j\ne F_l$, and $|F_l|<n/2+1$ ?

I'm hoping for a counterexample, and not for demonstration of a "yes" answer because this "yes" implies the Frankl Conjecture. Indeed, if $(G,\leq)=(\left\{G_i,i\in n+1\right\},\leq)$ with $G_n=0_G$ is a lattice and you take as $F_i=\left\{j\in i+1,\, G_j\leq G_i\right\}$, then $\left\{0,F_0,...,F_{n-1}\right\}$ is an intersection closed family (that satisfies the condition $(*)$ up to a fine indexation of elements in $G$. $(\mathcal F,\subset)$ is then a lattice that is isomorphic to $G$. If the answer to the question is "yes", then you will find some meet irréductible $G_i$ with at most $n/2$ lower elements. (The Frankl conjecture is known to be equivalent to the existence of such a $G_i$ in any lattice that cardinality is $n$.)

So the question seems much stronger than Frankl conjecture. So I expect it to "fail", never the less, there is an interesting intermediary question that can be asked if answer to main question is found to be "no".

Question 2

If $(H,\leq)$ is a partial order, does there exists, $h\in H$, such that $h$ is not the greater lower bound of any pair included in $H\setminus\left\{h\right\}$, (let's say that $H$ is then "quasi-irreductible") and such that $|\left\{x\in H, x\leq h\right\}|<|H|/2$.

If $H$ is a lattice, then "quasi-irreductible" and "irreductible" is the same thing.

But answer 2, in the general case, would not be "yes" , if we replace "any pair" by "any subset" :

Take any set $M$ of cardinality $n$ that elements are subset of $4n$ and such that each $m\in M$ has cardinality $3n$ suppose also that the smallest intersection closed set that contain $M$ also contain $S$ the set of singleton of $4n$. (it is not hard to build such a set if $n$ is big enought let's say $n>4$ for example). Now consider $M'=M\cup S$, than each member of $M'$ is a quasi-irreductible and the question 2 is yes in this trivial case, because any element of $S\subset M$' has zero member that are smaller then them, witch is less then $|M'|/2$. But if your definition of $h$ irreductible is "$h$ is not the greater lower bound of some subset of $M'\setminus\left\{h\right\}$" than the irreductible are exactly elements of $M$, each one of them has cardinality $3n$ and then contain $3n$ members of $S\subset M'$, and now... $3n>|M'|/2=(4n+n)/2$.

  • 1
    $\begingroup$ The very first notation is very confusing. $\endgroup$ Jan 27 '19 at 23:18
  • $\begingroup$ The question is : if M is a triangular matrix with only "zero" and "one" in entry and with "one" on the diagonal, is there a row that is not the "conjonction" of two other rows, and that has at most n/2 "ones" . (the conjonction of two rows, is the multiplication coordonate by coordonate, of the two rows, like a multiplication in a ring product, that the two rows can be seen in : ex : the conjonction of 11001 and 10101 is 10001) $\endgroup$
    – jcdornano
    Jan 30 '19 at 11:54
  • $\begingroup$ If we ask the triangular matrix to be the adjacency matrix of a lattice (up to the 0 rows and the zero column) we get the Frankl conjecture. Actually I forget that i put this post, and then i posted about a year later , i think, the "matrix version" that i just wrote in the upper comment. I will give the link of this (then redondant ) post, as soon as I get a PC (i'm now on the telephone) $\endgroup$
    – jcdornano
    Jan 30 '19 at 12:01

This is a fairly sorry counterexample to the main conjecture, but I do think it is one. Take $n=2$, $F_{0}=\{0\}$, $F_{1}=\{1\}$. Then neither $0$ or $1$ satisfy the conditions you're looking for on $l$.

Similarly, for Question 2, in a two-element anti-chain with $x\leq x$, $y\leq y$, but no relation between the two, there's no such $h$.

  • $\begingroup$ thank you, I made a mistake that I'm going to correct,the inequality is supposed to be large (like in Frankl conjecture). Sorry again. $\endgroup$
    – jcdornano
    Jan 30 '19 at 11:44
  • $\begingroup$ I added 1 to the right member wich is at least weaker than the large inequality (but in fact equivalent) $\endgroup$
    – jcdornano
    Jan 30 '19 at 11:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.