1
$\begingroup$

Let $a_n$ and $b_n$ be two different expressions in natural $n$ with values in the set of all nonnegative integers such that we have the identity $a_n=b_n$ for all $n$. As a simplest example, we may consider $a_n:=\sum_{k=0}^n\binom nk$ and $b_n:=2^n$.

Is there an automated, algorithmic construction of bijective proofs of broad enough classes of such identities?

Given $a_n$ and $b_n$, such an algorithm would parse the expressions $a_n$ and $b_n$ and then (i) construct finite sets $A_n$ and $B_n$, (ii) construct a bijection from $A_n$ onto $B_n$, and (iii) show that the cardinalities of $A_n$ and $B_n$ are $a_n$ and $b_n$, respectively.

Remark: To avoid undesirable, trivial constructions such as the one mentioned in a comment by Sam Hopkins, it should be required that the expressions for the sets $A_n$ and $B_n$ not contain the expressions $a_n$ and $b_n$ as terms.

Weaker versions of the above question are obtained by asking the steps (i)–(iii) to be done only for all small enough values of $n$, possibly followed by a conjecture for all $n$.

The above question is, of course, somewhat vague, because it is hard to attach in advance a definite meaning to the phrase "broad enough classes".

In 2017, Timothy Chow asked a related question Automated search for bijective proofs, which was well thought-out and well presented. The main difference is that there sets $A_n$ and $B_n$ were given.

Even with $A_n$ and $B_n$ given, the problem appears very difficult. However, there may have been progress on such problems since 2017.

$\endgroup$
4
  • $\begingroup$ You have to be really careful how you phrase things. Any proof at all the $a_n=b_n$ would technically fit your (i), (ii), (iii), with $A_n = \{1,2,\ldots,a_n\}$ and $B_n = \{1,2,\ldots,b_n\}$. $\endgroup$ Commented Dec 19, 2021 at 16:34
  • $\begingroup$ @SamHopkins : Thank you for your comment. I have added a remark to address this issue. Do you think this will be enough? $\endgroup$ Commented Dec 19, 2021 at 16:57
  • $\begingroup$ I think it is a more subtle question than that. See the discussion at mathoverflow.net/questions/323779/… $\endgroup$ Commented Dec 19, 2021 at 17:05
  • $\begingroup$ @SamHopkins : Thank you for your further suggestion. Can you give a specific objection showing that the added remark is still not enough? Please note that the question is only about some selected, delineated classes of identities. $\endgroup$ Commented Dec 19, 2021 at 17:27

0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.