3

I was playing with Pascal's triangle and noticed an interesting pattern: most numbers (except 1) appear only once (e.g., 2, 3, 4) or twice (e.g., $6 = \binom{4}{2} = \binom{6}{1}$), when excluding duplicates within the same row.

To investigate this further, I wrote a simple Python script to count how many times each number appears in Pascal's triangle:

import math
import matplotlib.pyplot as plt
import numpy as np

a = [] b = [] for n in range(2, 10000): count = 0 for k in range(n + 1): for j in range(math.floor(k / 2) + 1): if math.comb(k, j) > n: break if math.comb(k, j) == n: count += 1 b.append(count) a.append(n)

plt.plot(np.array(a), np.array(b)) plt.show()

enter image description here

The first number that appears three times is $210$:
$$ 210 = \binom{10}{4} = \binom{21}{2} = \binom{210}{1} $$ The first number to appear four times is (3003):
$$ 3003 = \binom{14}{6} = \binom{15}{5} = \binom{78}{2} = \binom{3003}{1} $$ No number is repeated five times in (1,30000)

I conjecture that for any $ N \in \mathbb{N} $, there exists a number $ n $ that appears exactly $ N $ times in Pascal’s triangle. However, I have been unable to prove or disprove this.

  1. Is this conjecture true?
  2. If false, what is the maximum number of times a number can appear in Pascal’s triangle?

Additionally, Assuming this conjecture holds I am curious about the sequence $ a_n $, where $ a_n $ is the first number to appear $ n $ times in Pascal’s triangle. The known values seem to be:
$$ 2, 6, 210, 3003, \dots $$ This sequence does not appear in the OEIS, and finding a closed formula seems difficult due to the chaotic nature of the sequence.

  1. Can we find an upper bound on $ a_n $? Empirically, $ a_n $ seems to grow rapidly. can we find a function $f$ such that $a_n =O(f(n))$
pie
  • 8,483
  • Typo: you meant to write "The first number to appear four times is $3003$" not five times. But I don't understand how you are counting. Doesn't $210=\binom {210}{209}$ for example? – lulu Feb 09 '25 at 19:04
  • 1
    Are you sure about your python script? https://en.wikipedia.org/wiki/Singmaster%27s_conjecture#Elementary_examples has slightly different numbers, including 120 in the same category as 210 – DrM Feb 09 '25 at 19:08
  • @lulu: "when excluding duplicates within the same row." So in other words, the number of different rows that it appears in. – Nate Eldredge Feb 09 '25 at 19:08
  • @lulu I exclude repetition in the same row of pascal triangle – pie Feb 09 '25 at 19:09
  • @NateEldredge Ah, thank you. In any case, the duplicate covers the multiplicity question. – lulu Feb 09 '25 at 19:09
  • @lulu This question has more question than the other question in fact what I am intersted about is the sequence $2,6,210,3003,$ and finding a function $f$ such that $\lim_{n\to\infty}\frac{a_n}{f(n)}< M\in \mathbb {R}$ – pie Feb 09 '25 at 19:12
  • and once you adjust to include 120, https://oeis.org/A062527 – DrM Feb 09 '25 at 19:12
  • 1
    Well...this is the problem with asking multiple questions in a post. Some of your questions are "answered" by the duplicate (that is, related to a known open conjecture) but other questions may not be. But...since the multiplicity is conjectured to be bounded, is there any question left? You can just take $f(n)=1$. But, again, the conjecture is open. – lulu Feb 09 '25 at 19:13
  • @DrM ahaa, I made a mistake then. When I wrote it with 210 it didn't appear in OEIS. – pie Feb 09 '25 at 19:14
  • @lulu No, I retracted my reopen-vote. – pie Feb 09 '25 at 19:15
  • To be thorough, as the conjecture is open, one might ask whether there is some known function, even if it tends to infinity, but for that I suggest going through the literature on Singmaster's conjecture to see what partial results are available. – lulu Feb 09 '25 at 19:17
  • @lulu I think even if such thing exist it would be too advanced for me. – pie Feb 09 '25 at 19:20
  • Note the result of Singmaster cited in the Wikipedia page ($N(a)= O(\log a)$ shows that if your $a_n$ exists for all $n$ (which is what you conjectured as true and what Singmaster conjectured as false), then it grows at least exponentially fast: $a_n \ge e^{n/C}$ for some $C$. Some sharper bounds are given too but a little harder to translate into bounds on $a_n$. – Nate Eldredge Feb 09 '25 at 22:57

0 Answers0