Re: Re: ArcCos[]
- To: mathgroup at smc.vnet.net
- Subject: [mg24798] Re: [mg24793] Re: [mg24730] ArcCos[]
- From: Andrzej Kozlowski <andrzej at tuins.ac.jp>
- Date: Sun, 13 Aug 2000 03:16:38 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
I have had a further exchange of messages with Gianluca about this topic
which lead to conclude that the problem is simpler than I assumed. Below is
an extract from a message I sent to Gianluca:
I think now that actually it would be possible to implement this, (although
it may not be worth the effort). What one would really need to to is to find
a mathematical characterization of the minimal polynomials of algebraic
numbers of the form Cos[r*Pi] for rational r. This should not be too hard.
Then given a radical expression like Sqrt[2 + Sqrt[2]]/2 one simply applies:
In[51]:=
RootReduce[Sqrt[2 + Sqrt[2]]/2]
Out[51]=
2 4
Root[1 - 8 #1 + 8 #1 & , 4]
and then one "looks" at the minimal polynomial. For example, look at my
other example (Cos[Pi/15]):
In[64]:=
RootReduce[((-1 + Sqrt[5]))/8 + (Sqrt[(3*(5 + Sqrt[5]))/2])/4]
Out[64]=
2 3 4
Root[1 - 8 #1 - 16 #1 + 8 #1 + 16 #1 & , 4]
One can see some very clear regularity, suggesting certain conjectures. By
comparison, taking a radical "at random":
In[57]:=
RootReduce[Sqrt[(1 + Sqrt[7])/4]]
Out[57]=
2 4
Root[-3 - 4 #1 + 8 #1 & , 2]
This clearly looks quite different. So I now believe a rigorous mathematical
solution is possible and very likely well known to number theorists, but
perhaps no sufficiently important for Mathematica developers to bother
about.
Andrzej
on 8/10/00 6:33 AM, Andrzej Kozlowski at andrzej at tuins.ac.jp wrote:
> on 8/9/00 8:31 AM, Gianluca Gorni at gorni at dimi.uniud.it wrote:
>
>>
>> Hello!
>>
>> I have just come across one more example that shows how there is
>> room for improving Mathematica's trig functions.
>>
>> With Mathematica 4:
>>
>> v = Cos[ Pi/8 ] // FunctionExpand gives Sqrt[2 + Sqrt[2]]/2
>>
>> Still, neither
>>
>> ArcCos[v] // FunctionExpand nor ArcCos[v] // FullSimplify
>>
>> give Pi/8, as I would expect, but just
>>
>> ArcSec[2/Sqrt[2 + Sqrt[2]]]
>>
>> %%%%%%%%%%%%%%%
>>
>> An unrelated problem: the following instructions consistently crash
>> my Mac Mathematica 4 kernel:
>>
>> a = Root[-t + 2*#1 + 2*t^2*#1 + #1^3 & , 1];
>> b = D[a, t];
>> Solve[b == 0, t]
>>
>> %%%%%%%%%%%%%%%
>>
>> Best regards,
>>
>> Gianluca Gorni
>
> I am not an expert on this sort of thing, but "mathematical common sense"
> suggest to me that this may not be easy. Let's consider carefully the
> problem of finding a "radical" expression for Cos[Pi/8] (or indeed any
> Tr[Pi*m] where Tr is a trigonometric function and m a rational). Although
> they do not look like it at first sight these expressions are in fact
> algebraic numbers. The point is that Cos[Pi/8] is just the real part of the
> cyclotomic number Cos[2*Pi/16]+I*Sin[2*Pi/16] which is just Root[#^16 - 1 &,
> 16] as you can see from:
>
> In[54]:=
> (Root[#^16 - 1 &, 16] - Cos[2*Pi/16] - I*Sin[2*Pi/16]) // FullSimplify
> Out[54]=
> 0
>
> So we can write Cos[Pi/8] as 1/2(Root[#^16 - 1 &, 16] + 1/Root[#^16 - 1 &,
> 16]), which is, of course, an algebraic number. Using RootReduce we can get
>
> In[58]:=
> RootReduce[1/2(Root[#^16 - 1 &, 16] + 1/Root[#^16 - 1 &, 16])]
> Out[58]=
> 2 4
> Root[1 - 8 #1 + 8 #1 & , 4]
>
> and then
>
> In[60]:=
> ToRadicals[%] // FullSimplify
> Out[60]=
> Sqrt[2 + Sqrt[2]]
> -----------------
> 2
>
> Of course in this case we could have much easier got this expression
> starting from the known values of the trigonometric functions of Pi/4 and
> then using half angle formulas. But my point is that there is a procedure
> that can be tried in general in such cases.
>
> However, conversely: given a radical expresion e.g. v=((-1 + Sqrt[5]))/8 +
> (Sqrt[(3*(5 + Sqrt[5]))/2])/4 how do you go about deciding if it is the real
> or complex part of a cyclotomic number? In fact, v= Cos[Pi/15] but I
> doubt that there is any algorithm which has a reasonable chcance of
> determining if an arbitrary radical expression (with value between -1 and 1,
> say) is the value of a trigonometric function of some rational multiple of
> Pi.
> (If I am wrong I would like to hear about this).
>
>