5

I am trying to calculate the $n^{\text{th}}$ term of the following polynomial:

$$\, _2F_1\left(-n,n+3;\frac{3}{2};x\right)$$

To do this I calculate:

c[k_] = SeriesCoefficient[
    Hypergeometric2F1[-n, n+3, 3/2, x], {x, 0, k},
    Assumptions -> k >= 0
];

and get:

c[k] //TeXForm

$\frac{\sqrt{\pi } (k-n-1)! (k+n+2)!}{2 k! \left(k+\frac{1}{2}\right)! (-n-1)! (n+2)!}$

and the problem is when I try to calculate for $n=10$:

Block[{n = 10}, Sum[c[k] x^k, {k, 0, 10}]]

Infinity::indet: Indeterminate expression (0 2 Sqrt[π] ComplexInfinity)/Sqrt[π] encountered.

Infinity::indet: Indeterminate expression (0 4 Sqrt[π] ComplexInfinity)/(3 Sqrt[π]) encountered.

Infinity::indet: Indeterminate expression (0 8 Sqrt[π] ComplexInfinity)/(15 Sqrt[π]) encountered.

General::stop: Further output of Infinity::indet will be suppressed during this calculation.

Indeterminate

I get error messages and an incorrect answer. The correct result is:

Hypergeometric2F1[-n, n+3, 3/2, x] /. n->10

1/33 (33 - 2860 x + 72072 x^2 - 823680 x^3 + 5125120 x^4 - 19009536 x^5 + 43868160 x^6 - 63504384 x^7 + 56033280 x^8 - 27525120 x^9 + 5767168 x^10)

Other manifestations of problems with c:

Block[{n=10}, c[k]]
Block[{n=10}, c[5]]

0

Infinity::indet: Indeterminate expression -((128 0 64 316234143225 Sqrt[π] Sqrt[π] ComplexInfinity)/(135135 10395 4096 Sqrt[π] Sqrt[π])) encountered.

Indeterminate

c does not give a useful symbolic result for the $k^{\text{th}}$ term of the series.

I try to use assumptions but it does not help.

m_goldberg
  • 107,779
  • 16
  • 103
  • 257
Clerk
  • 65
  • 3
  • 1
    It will be impossible to determine why you are getting an error without code. – N.J.Evans Aug 30 '18 at 12:30
  • please how to paste the code i copy from mathematica +crtl K but do not work – Clerk Aug 30 '18 at 12:47
  • I put a code usinf crtl +k + Latex code please how it do ?? – Clerk Aug 30 '18 at 12:58
  • You should be able to copy it straight out of an MMA notebook by selecting it and using copy and paste. Or you can type it directly in here, though there is always a chance that you introduce new typos. If you inclose what you type in backticks it will show up as code e.g. `f[x_]:=x^2` shows up as: f[x_]:=x^2 – N.J.Evans Aug 30 '18 at 13:32
  • You can read more about markdown editing here, if you're interested. – N.J.Evans Aug 30 '18 at 13:33
  • 1
    The code corresponding to the question gives me the correct answer with no error messages. I suggest saving your notebook (without any extraneous material), restarting Mathematica, and then executing the notebook. If that does not work, check your code for errors. – bbgodfrey Aug 30 '18 at 13:49
  • To be specific, SeriesCoefficient[Hypergeometric2F1[-n, n + 3, 3/2, x], {x, 0, k}] followed by Sum[% x^k, {k, 0, n}] gives the correct answer. – bbgodfrey Aug 30 '18 at 13:51
  • bbodfrey how it is possible in the denominator Gamma[-1-n] when n=10 gives infinity I get the error Infinity::indet: Indeterminate expression (0 256 Sqrt[[Pi]] Sqrt[[Pi]] ComplexInfinity – Clerk Aug 30 '18 at 14:36
  • for example n = 10 Sum[(Sqrt[[Pi]] (-1 + k - n)! (2 + k + n)!)/( 2 k! (1/2 + k)! (-1 - n)! (2 + n)!) x^k, {k, 0, n}] gives indeterminate please help – Clerk Aug 30 '18 at 14:43
  • 1
    Do the Sum first (with n left undefined, then substituten->10` in the result. This will avoid the indeterminate forms. – Daniel Lichtblau Aug 30 '18 at 17:04
  • Thanks Daniel but this not rsolve the problem – Clerk Aug 30 '18 at 17:15
  • 1
    If the comment by Daniel ( %/.n->10 ) doesn't solve your problem, when using a fresh kernel as suggested by bbgodfrey, then it's not clear what you are asking. Please [edit] your question to clarify. – rhermans Aug 31 '18 at 07:51

2 Answers2

8

It seems that the output generated by SeriesCoefficient is difficult for Mathematica to simplify into a version that can evaluate properly for integer n. So, I recommend using the new symbolic order derivatives introduced in M11.1:

c[k_] = Assuming[
    n>=1,
    Simplify @ D[Hypergeometric2F1[-n, n+3, 3/2, x], {x, k}]/k! /. x->0
]

(Pochhammer[-n, k] Pochhammer[3 + n, k])/(k! Pochhammer[3/2, k])

Note that this version of the symbolic coefficient of the series evaluates correctly for explicit values of n and k:

Block[{n=10}, c[5]]

-6336512/11

Let's check:

r1 = Block[{n=10}, Sum[c[k] x^k, {k, 0, n}]];
r1 // TeXForm

$\frac{524288 x^{10}}{3}-\frac{9175040 x^9}{11}+\frac{18677760 x^8}{11}-\frac{21168128 x^7}{11}+\frac{14622720 x^6}{11}-\frac{6336512 x^5}{11}+\frac{465920 x^4}{3}-24960 x^3+2184 x^2-\frac{260 x}{3}+1$

Compare to the exact answer:

r2 = Block[{n=10}, Expand @ Hypergeometric2F1[-n,n+3,3/2,x]];
r2 // TeXForm

$\frac{524288 x^{10}}{3}-\frac{9175040 x^9}{11}+\frac{18677760 x^8}{11}-\frac{21168128 x^7}{11}+\frac{14622720 x^6}{11}-\frac{6336512 x^5}{11}+\frac{465920 x^4}{3}-24960 x^3+2184 x^2-\frac{260 x}{3}+1$

They are the same:

r1 === r2

True

Addendum

The OP asks in a comment about a different hypergeometric function argument:

c[k_]=Assuming[
    n>=1,
    Simplify @ D[Hypergeometric2F1[3/2+n, -(3/2)-n, 3/2, x], {x, k}]/k! /. x->0
];

r1 = Block[{n=10}, Sum[c[k] x^k, {k, 0, n}]];
r1 //TeXForm

$\frac{515830463005 x^{10}}{262144}-\frac{264205846905 x^9}{65536}+\frac{165491574435 x^8}{32768}-\frac{8448518815 x^7}{2048}+\frac{2304141495 x^6}{1024}-\frac{2304141495 x^5}{2816}+\frac{24775715 x^4}{128}-\frac{452295 x^3}{16}+\frac{18515 x^2}{8}-\frac{529 x}{6}+1$

r2 = Block[{n=10}, Hypergeometric2F1[3/2+n, -(3/2)-n, 3/2, x] //Expand];
r2 //TeXForm

$-\frac{524288}{3} \sqrt{1-x} x^{11}+\frac{33292288}{33} \sqrt{1-x} x^{10}-\frac{27852800}{11} \sqrt{1-x} x^9+\frac{39845888}{11} \sqrt{1-x} x^8-\frac{35790848}{11} \sqrt{1-x} x^7+\frac{20959232}{11} \sqrt{1-x} x^6-\frac{24134656}{33} \sqrt{1-x} x^5+\frac{540800}{3} \sqrt{1-x} x^4-27144 \sqrt{1-x} x^3+\frac{6812}{3} \sqrt{1-x} x^2-\frac{263}{3} \sqrt{1-x} x+\sqrt{1-x}$

The difference between them is that r1 is a series approximation of r2. When r2 is not a degree 10 polynomial, than the two expressions will not be the same. Instead compare r1 with the series approximation of r2:

r2 + O[x]^11 //TeXForm

$1-\frac{529 x}{6}+\frac{18515 x^2}{8}-\frac{452295 x^3}{16}+\frac{24775715 x^4}{128}-\frac{2304141495 x^5}{2816}+\frac{2304141495 x^6}{1024}-\frac{8448518815 x^7}{2048}+\frac{165491574435 x^8}{32768}-\frac{264205846905 x^9}{65536}+\frac{515830463005 x^{10}}{262144}+O\left(x^{11}\right)$

Carl Woll
  • 130,679
  • 6
  • 243
  • 355
  • Thanks Carlk for your help it is works when n is neative interger but i try with n=3/2-n and do not work for example c[k_] = Assuming[n >= 1, Simplify@D[Hypergeometric2F1[3/2 + n, -(3/2) - n, 3/2, x], {x, k}]/ k! /. x -> 0] r1 = Block[{n = 10}, Sum[c[k] x^k, {k, 0, n}]]; r1 r2 = Block[{n = 10}, Hypergeometric2F1[3/2 + n, -(3/2) - n, 3/2, x] // Expand]; r2 gives different result check it if you will ,(sorry i could paste de code cause i paste from mathematica +Crtl K but do not work for me thanks anyway – Clerk Aug 31 '18 at 09:59
  • Thanks Carlk now it is very clear – Clerk Aug 31 '18 at 13:42
5

Instead = write := at function c[k_]

$Version
(* "11.3.0 for Microsoft Windows (64-bit) (March 7, 2018)"*)

ClearAll["Global`*"]; Remove["Global`*"];(* Clears the kernel *)

c[k_] := SeriesCoefficient[Hypergeometric2F1[-n, n + 3, 3/2, x], {x, 0, k}]
Block[{n = 10}, Sum[c[k] x^k, {k, 0, 10}]]

$\frac{524288 x^{10}}{3}-\frac{9175040 x^9}{11}+\frac{18677760 x^8}{11}-\frac{21168128 x^7}{11}+\frac{14622720 x^6}{11}-\frac{6336512 x^5}{11}+\frac{465920 x^4}{3}-24960 x^3+2184 x^2-\frac{260 x}{3}+1$

Mariusz Iwaniuk
  • 13,841
  • 1
  • 25
  • 41