Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

How do they do that?

7 views
Skip to first unread message

Mikito Harakiri

unread,
Dec 31, 2003, 5:59:50 PM12/31/03
to
> simplify((sqrt(5)+2)^(1/3)-(sqrt(5)-2)^(1/3));

1


I enabled trace level 50, or so, and was overwhelmed with information. It
seems like Maple tries to find a numerical approximation, first. Why
numerical approximation matters?

(I have read about this problem that roots in this expression are resistant
to any straightforward application of power operation; one have to consider
a cubic equation root of which this expression is).

Richard Fateman

unread,
Dec 31, 2003, 9:12:09 PM12/31/03
to
Maybe they started by reading the literature, see, for
example,
http://citeseer.nj.nec.com/landau93note.html

As for why you might want a numerical approximation,
sometimes you want to know if expression > 0.
If you can determine that expression is between 1.5 and 1.6
but you don't know even the next digit, you can still answer
the question.

RJF

Robert Israel

unread,
Dec 31, 2003, 9:25:08 PM12/31/03
to
In article <HFIIb.6$Mf4...@news.oracle.com>,

Mikito Harakiri <mikha...@iahu.com> wrote:
>> simplify((sqrt(5)+2)^(1/3)-(sqrt(5)-2)^(1/3));

> 1


>I enabled trace level 50, or so, and was overwhelmed with information. It
>seems like Maple tries to find a numerical approximation, first. Why
>numerical approximation matters?

I suspect that the fact that sqrt(5)-2 > 0 may be important here, in
deciding which branch of some multivalued function to choose. This
can be tricky.

Somewhat more generally, I believe
(sqrt(b^3+a^2)+a)^(1/3) - (sqrt(b^3+a^2)-a)^(1/3)
is one of the roots of z^3 + 3 b z - 2 a if b > 0 and a is real,
but not necessarily in other cases.

Robert Israel isr...@math.ubc.ca
Department of Mathematics http://www.math.ubc.ca/~israel
University of British Columbia
Vancouver, BC, Canada V6T 1Z2

steve_H

unread,
Dec 31, 2003, 10:08:16 PM12/31/03
to
"Mikito Harakiri" <mikha...@iahu.com> wrote in message news:<HFIIb.6$Mf4...@news.oracle.com>...
> > simplify((sqrt(5)+2)^(1/3)-(sqrt(5)-2)^(1/3));
>
> 1
>
>

This is funny, I can get 4/3 not 1. here is how.


using the relation,


exp( (1/3)*ln(x) ) = x^1/3

then write ( sqrt(5)+2 )^1/3 ---> exp( (1/3)*ln( sqrt(5)+2 ) )

then let x=5, and use series expansion for

exp( (1/3)*ln( sqrt(x)+2 ) )

and

exp( (1/3)*ln( sqrt(x)-2 ) )

and subtract the difference. here it is:


> a:=series( (1/3)*exp(ln(sqrt(x)+2)),x);

1 (1/2) 2
a := - x + -
3 3

> b:=series( (1/3)*exp(ln(sqrt(x)-2)),x);


1 (1/2) 2
b := - x - -
3 3

> a-b;

4
-
3

and not 1 . may be I made a mistake somewhere.

Carl Devore

unread,
Jan 1, 2004, 6:31:04 AM1/1/04
to
On 31 Dec 2003, steve_H wrote:
> "Mikito Harakiri" <mikha...@iahu.com> wrote in message news:<HFIIb.6$Mf4...@news.oracle.com>...
> > > simplify((sqrt(5)+2)^(1/3)-(sqrt(5)-2)^(1/3));
> > 1
> exp( (1/3)*ln( sqrt(x)+2 ) )
> > a:=series( (1/3)*exp(ln(sqrt(x)+2)),x);

> and not 1 . may be I made a mistake somewhere.

Look at the different placement of the 1/3 in the two lines above.

0 new messages