Title: The Number of Ways of Expressing t as a Binomial Coefficient
1The Number of Ways of Expressing t as a Binomial
Coefficient
- By Daniel Kane
- January 7, 2007
2An Interesting Note
10 is both a triangular number and a tetrahedral
number.
3Why does this hold?
Is it equal to any other binomial coefficient?
4Are there any non-trivial relations?
Yields a Pell Equation
with Solutions
5Stating the Problem More Precisely
- Studied by Singmaster in 2 (1971)
- N(3003) 8
- N(t) 6 for infinitely many t
- Conj N(t) O(1)
6What are Reasonable Bounds?
- Singmaster showed in 2 that
- Consider n gt 2m (we do this from now on)
7An improvement
- In 3 Erdös et al. proved that
- Prime in (n-n5/8, n) for n gtgt 1.
- Split into cases based on n gt (log t)6/5
8For n gt (log t)6/5
9For n lt (log t)6/5
- Approximation yields m gt n5/8
- 9 prime, P, s.t. n-m lt P lt n
- P divides t
- Pick largest N, all others satisfy P lt n lt N
- At most M solutions
- M O(N5/8) O((log t)3/4)
- Using the strongest conjectures on gaps between
primes could give
10Another way to handle this case
- Consider all solutions
- Order so n1 gt n2 gt gt nk
- m1 lt m2 lt lt mk,
- As n decreases and m increases, the effect of
changing n decreases and the effect of changing m
increases. - (ni-ni1)/(mi1-mi) increasing.
- These fractions are distinct
11Another way (continued)
- Differences of (ni , mi) are distinct
- Only O(s2) have ?n, ?m lt s
- Total Change in n-m gt ck3/2
- k3/2 O((log t)6/5)
- k O ((log t)4/5).
12What did we do?
- n convex function of m
- Lattice points on graph of ANY convex function
- Idea Consider higher order derivatives.
13Problems to Overcome
- How do we use the derivative data?
- How do we obtain the derivative data?
14Using Data, a Lemma
- Lemma If f and g are Cn-functions so that f(x)
g(x) at x1 lt x2 lt lt xn1, then - f(n)(y) g(n)(y) for some y2 (x1 , xn1)
- Generalization of Rolles Theorem
- Proof by induction Rolles Theorem
15Using Data, Proof of Lemma
16Using Derivative Data (Continued)
- Consider function f, f(mi) ni
- g polynomial interpolation of f at m1, m2, ,
mk1 - g(k) constant
- f(k) is this constant at some point
17Using Derivative Data (cont.)
18Using Derivative Data (cont.)
- kth derivate of f small but non-zero
- Fit polynomial to k1 points separated by S
- ) kth derivative over k! either 0 or more than
S-k(k1)/2.
19Derivative Data
- Define
- Make smooth using ?
- Estimate with Sterlings approximation
20Estimating Derivatives
- Main Term Derivatives of
- log t ?(z1) / z and take exp of power
series - (z - 1) / 2 is easy
- Error term Cauchy Integral Formula
21A Useful Parameter
22Split into Cases
- ? lt 1.15
- 1.15 lt ? lt log log t/(24 log log log t)
- log log t/(24 log log log t)lt ? lt(log log t)4
- (log log t)4 lt ?
23Case 1 ? lt 1.15
- Already covered
- O((log t)3/4) solutions
24Case 2 1.15 lt ? lt log log t/(24 log log log t)
- Set k (log log t)/(12 log log log t)
- Technical conditions satisfied
- k1 adjacent solutions mi of separation S
25Case 3 log log t/(24 log log log t)lt ? lt(log
log t)4
- We need a slightly better analysis to bound
LCM over all sequences of k distinct ri? 0, ri
lt S
26Bounding B
- Count multiples of each prime
- pn divides at most min(k , 2S/pn) ris
- Use Prime Number Theorem
27Using Bound
- k 2?
- Technical Conditions Satisfied
28Case 4 (log log t)4 lt ?
29Conclusions
- Know where to look to tighten this bound
- Can use technique for other problems
30References
- 1 D. Kane, On the Number of Representations of
t as a Binomial Coefficient, Integers Electronic
Journal of Combinatorial Number Theory, 4,
(2004), A07, pp. 1-10. - 2 D. Singmaster, How often Does an Integer
Occur as a Binomial Coefficient?, American
Mathematical Monthly, 78, (1971) 385-386 - 3 H. L. Abbott, P. Erdös and D. Hanson, On the
number of times an integer occurs as a binomial
coefficient, American Mathematical Monthly, 81,
(1974) 256-261.