Re: simple Sin

• To: mathgroup at smc.vnet.net
• Subject: [mg89924] Re: [mg89859] simple Sin
• From: Andrzej Kozlowski <akoz at mimuw.edu.pl>
• Date: Tue, 24 Jun 2008 03:32:02 -0400 (EDT)
• References: <200806230646.CAA00151@smc.vnet.net>

```In order to get a higher order Taylor expansion you have to use a
higher order numerical ODE solver. (The default uses finite
differences of order one.)

NDSolve[{Derivative[2][y][t] + y[t] == 0, Derivative[1][y][0] == 1,
y[0] == 0}, y,
{t, 0, 2*Pi}, Method -> {"ExplicitRungeKutta", "DifferenceOrder" -
> 6}];

sin[u_] = y[u] /. First[%];

N[Normal[Series[{sin[t], Sin[t]}, {t, 0, 5}]]]

{0.008325597762864565*t^5 + 8.18663497983204*^-7*t^4 -
0.16666669041252172*t^3 +
3.469446951953614*^-18*t^2 + 1.*t, 0.008333333333333333*t^5 -
0.16666666666666666*t^3 + t}

You still get terms of degree 4 and 2 but with very small
coefficients. You can't possibly expect to get these to be exactly 0
with a purely numerical method!

Andrzej Kozlowski

On 23 Jun 2008, at 15:46, Narasimham wrote:

> It is surprising a bit, by numerical computation Sin[ t] has a
> different series representation, is not even an odd function of t !
> Looks like has a different chemistry.
>
> NDSolve[{y''[t] + y[t] == 0, y'[0] == 1, y[0] == 0 }, y, {t, 0, 2 Pi}]
> sin[u_] = y[u] /. First[%]
> Plot[sin[t] - Sin[t], {t, 0, 2 Pi}]
>
> Series[{sin[t], Sin[t]}, {t, 0, 8}]
>
> Also, why do we not get an expansion of  sin as Series[Im[Exp[I*t]],
> {t, 0, 8}] ?
>
> TIA
>
> Narasimham
>

```

• References:
• simple Sin
• From: Narasimham <mathma18@hotmail.com>
• Prev by Date: Re: Solving a Sum
• Next by Date: Re: Mathematica performance improvements
• Previous by thread: Re: simple Sin
• Next by thread: Re: simple Sin