login
Matrix inverse of the Narayana triangle A001263.
8

%I #38 Jul 02 2022 21:47:48

%S 1,-1,1,2,-3,1,-7,12,-6,1,39,-70,40,-10,1,-321,585,-350,100,-15,1,

%T 3681,-6741,4095,-1225,210,-21,1,-56197,103068,-62916,19110,-3430,392,

%U -28,1,1102571,-2023092,1236816,-377496,68796,-8232,672,-36,1,-27036487,49615695,-30346380,9276120,-1698732,206388

%N Matrix inverse of the Narayana triangle A001263.

%C The first column is A103365. The second column is A103366. Row sums are all zeros (for n > 1). Absolute row sums form A103367.

%C Let E(y) = Sum_{n >= 0} y^n/(n!*(n+1)!) = (1/sqrt(y))*BesselI(1,2*sqrt(y)). Then this triangle is the generalized Riordan array (1/E(y), y) with respect to the sequence n!*(n+1)! as defined in Wang and Wang. - _Peter Bala_, Aug 07 2013

%H G. C. Greubel, <a href="/A103364/b103364.txt">Table of n, a(n) for the first 50 rows, flattened</a>

%H W. Wang and T. Wang, <a href="http://dx.doi.org/10.1016/j.disc.2007.12.037">Generalized Riordan array</a>, Discrete Mathematics, Vol. 308, No. 24, 6466-6500.

%F From _Peter Bala_, Aug 07 2013: (Start)

%F Let E(y) = Sum_{n >= 0} y^n/(n!*(n+1)!) = (1/sqrt(y))* BesselI(1,2*sqrt(y)). Generating function: E(x*y)/E(y) = 1 + (-1 + x)*y/(1!*2!) + (2 - 3*x + x^2)*y^2/(2!*3!) + (-7 + 12*x - 6*x^2 + x^3)*y^3/(3!*4!) + .... The n-th power of this array has a generating function E(x*y)/E(y)^n. In particular, the matrix inverse A001263 has a generating function E(y)*E(x*y).

%F Recurrence equation for the row polynomials: R(n,x) = x^n - Sum_{k = 0..n-1} 1/(n-k+1)*binomial(n,k)*binomial(n+1,k+1) *R(k,x) with initial value R(0,x) = 1.

%F Let alpha denote the root of E(x) = 0 that is smallest in absolute magnitude. Numerically, alpha = -3.67049 26605 ... = -(A115369/2)^2. It appears that for arbitrary complex x we have lim_{n->oo} R(n,x)/R(n,0) = E(alpha*x). Cf. A055133, A086646 and A104033.

%F A stronger result than pointwise convergence may hold: the convergence may be uniform on compact subsets of the complex plane. This would explain the observation that the real zeros of the polynomials R(n,x) seem to converge to the zeros of E(alpha*x) as n increases. Some numerical examples are given below. (End)

%F From _Werner Schulte_, Jan 04 2017: (Start)

%F T(n,k) = T(n-1,k-1)*n*(n-1)/(k*(k-1)) for 1 < k <= n;

%F T(n,k) = T(n+1-k,1)*A001263(n,k) for 1 <= k <= n;

%F Sum_{k=1..n} T(n,k)*A000108(n) = 1 for n > 0. (End)

%e Rows begin:

%e 1;

%e -1, 1;

%e 2, -3, 1;

%e -7, 12, -6, 1;

%e 39, -70, 40, -10, 1;

%e -321, 585, -350, 100, -15, 1;

%e 3681, -6741, 4095, -1225, 210, -21, 1;

%e -56197, 103068, -62916, 19110, -3430, 392, -28, 1;

%e 1102571, -2023092, 1236816, -377496, 68796, -8232, 672, -36, 1;

%e ...

%e From _Peter Bala_, Aug 09 2013: (Start)

%e The real zeros of the row polynomials R(n,x) appear to converge to zeros of E(alpha*x) as n increases, where alpha = -3.67049 26605 ... ( = -(A115369/2)^2).

%e Polynomial | Real zeros to 5 decimal places

%e = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

%e R(5,x) | 1, 3.57754, 3.81904

%e R(10,x) | 1, 3.35230, 7.07532, 9.14395

%e R(15,x) | 1, 3.35231, 7.04943, 12.09668, 15.96334

%e R(20,x) | 1, 3.35231, 7.04943, 12.09107, 18.47845, 24.35255

%e = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

%e Function |

%e = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

%e E(alpha*x) | 1, 3.35231, 7.04943, 12.09107, 18.47720, 26.20778, ...

%e Note: The n-th zero of E(alpha*x) may be calculated in Maple 17 using the instruction evalf( BesselJZeros(1,n)/ BesselJZeros(1,1))^2 ). (End)

%t T[n_, 1]:= Last[Table[(-1)^(n - 1)*(CoefficientList[Series[x/BesselJ[1, 2*x], {x, 0, 1000}], x])[[k]]*(n)!*(n - 1)!, {k, 1, 2*n - 1, 2}]]

%t T[n_, n_] := 1; T[2, 1] := -1; T[3, 1] := 2; T[n_, k_] := T[n, k] = T[n - 1, k - 1]*n*(n - 1)/(k*(k - 1)); Table[T[n, k], {n, 1, 50}, {k, 1, n}] // Flatten (* _G. C. Greubel_, Jan 04 2016 *)

%o (PARI) T(n,k)=if(n<k || k<1, 0, (matrix(n,n,m,j,binomial(m-1,j-1)*binomial(m,j-1)/j)^-1)[n,k])

%Y Cf. A000108, A001263, A055133, A086646, A103365, A103366, A103367, A104033.

%K sign,tabl

%O 1,4

%A _Paul D. Hanna_, Feb 02 2005