OFFSET
2,1
COMMENTS
For n = 1 the equation is trivially solved by any integer, therefore we only consider n >= 2.
If any x_k = 0, then all x_i must be zero, so (0, ..., 0) would be the only additional solution in nonnegative integers. This solution is not considered here.
A vector (1, ..., 1, x_n) can never be a solution for n > 1. The number of components different from 1 must be k+1 >= 2 <=> k >= 1.
It can be shown that no solution can have 2^k > n^2, cf. the Shiu paper. Therefore row lengths are floor(2 log_2(n)) = (2, 3, 4, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, ...) = A329202(n), n >= 2.
Row sums yield the total number of nontrivial solutions A328910(n), see there for more information.
T(n,k) is equal to |C_k(n)| in the Shiu paper, but some values given in the table on top of p. 803 are erroneous (pers. comm. from the author).
LINKS
David A. Corneth, Table of n, a(n) for n = 2..14645 (first 899 rows flattened)
Peter Shiu, On Erdös's Last Equation, Amer. Math. Monthly, 126 (2019), 802-808.
EXAMPLE
The table starts:
n : T(n,k), 1 <= k <= 2*log_2(n)
2 : 2 0
3 : 3 3 0
4 : 4 3 1 0
5 : 4 4 0 0
6 : 6 7 4 0 0
7 : 6 5 3 0 0
8 : 5 7 4 2 1 0
9 : 8 13 5 1 0 0
10 : 9 12 3 1 0 0
11 : 6 6 3 0 0 0
12 : 8 13 9 3 0 0 0
13 : 8 7 1 0 0 0 0
14 : 6 15 6 2 1 0 0
15 : 12 16 12 3 0 0 0
For n = 2 variables, we have the equation x1*x2 = 2*(x1 + x2) with positive integer solutions (3,6) and (4,4): Both have k+1 = 2 components > 1, i.e., k = 1.
For n = 3, we have T(3,1) = 3 solutions with k+1 = 2 components > 1, {(1, 4, 15), (1, 5, 9), (1, 6, 7)}, and T(3,2) = 3 with k+1 = 3 components > 1, {(2, 2, 12), (2, 3, 5), (3,3,3)}.
For n = 4 we have the 8 solutions (1, 1, 5, 28), (1, 1, 6, 16), (1, 1, 7, 12), (1, 1, 8, 10), (1, 2, 3, 12), (1, 2, 4, 7), (1, 3, 4, 4) and (2, 2, 2, 6). Four of them have k+1 = 2 components > 1, i.e., k = 1, whence T(4,1) = 4. Three have k+1 = 3 <=> k = 2, so T(4,2) = 3. One has k+1 = 4, so T(4,3) = 1.
For n = 5, the solutions are, omitting initial components x_i = 1: {(6, 45), (7, 25), (9, 15), (10, 13), (2, 3, 35), (2, 5, 9), (3, 3, 10), (3, 5, 5)}. Therefore T(5,1..4) = (4, 4, 0, 0).
For n = 6, the solutions are (omitting x_i = 1): {(7, 66), (8, 36), (9, 26), (10, 21), (11, 18), (12, 16), (2, 4, 27), (2, 5, 15), (2, 6, 11), (2, 7, 9), (3, 3, 18), (3, 4, 10), (3, 6, 6), (2, 2, 2, 24), (2, 2, 3, 9), (2, 2, 4, 6), (2, 3, 3, 5)}. Therefore T(6,1..5) = (6, 7, 4, 0, 0).
For n = 9, the 27 solutions are (omitting '1's): {(10, 153), (11, 81), (12, 57), (13, 45), (15, 33), (17, 27), (18, 25), (21, 21), (2, 5, 117), (2, 6, 42), (2, 7, 27), (2, 9, 17), (2, 12, 12), (3, 4, 39), (3, 5, 21), (3, 6, 15), (3, 7, 12), (3, 9, 9), (4, 4, 18), (5, 5, 9), (6, 6, 6), (2, 2, 3, 36), (2, 2, 6, 9), (2, 3, 3, 13), (3, 3, 3, 7), (3, 3, 4, 5), (2, 3, 3, 3, 3)}. Therefore T(9,1..6) = (8, 13, 5, 1, 0, 0).
PROG
(PARI) A328911(n, k, show=1)={if( k<min(exponent(n^2)+1, n), my(s=0, t=n*(n-k-1), d); forvec(x=vector(k, i, [2, n\(sqrt(2)-1)]), (d=vecprod(x)-n)>0 && (n*vecsum(x)+t)%d==0 && (n*vecsum(x)+t)\d >= x[k] && s++&& show&& printf("%d, ", concat(x, (n*vecsum(x)+t)\d)), 1); s)}
CROSSREFS
KEYWORD
nonn,tabf
AUTHOR
M. F. Hasler, Nov 07 2019
STATUS
approved