There is a topic called analytic functions of a matrix that could be applied 
here, if you represent the quaternion a0 + a1 i + b0 j + b1 k by the matrix

Q =: 2 2 $ a , b, (- + b), - + a

where a means a0 j. a1  and b means b0 j. b1

When Q has distinct eigenvalues e0 and e1 with corresponding eigenvectors v0 
and 
v1, then

f(Q) = V mp (f E) mp %. V

when f is an analytic function, V is the matrix v0 ,. v1 and E is the diagonal 
matrix 2 2 % e0 , 0, 0, e1 and f E is the diagonal matrix 2 2 $ (f e0),0,0,f e1 
.  Function f has to be analytic in neighborhoods of e0 and e1.  mp is the 
matrix product +/ . *

Analytic functions include monadic ^ and ^. and functions like those found on 
the dictionary page for o. .  They make this topic about as hard as a typical 
engineering course in complex variables for which I recommend Fundamentals of 
Complex Analysis by E. B. Saff and A. D. Snider, Third Edition, Prentice Hall 
2003.

You can have x^y when quaternion y is a complex number but thinking of z^y as 
an 
analytic function of z has the difficulties presented in a complex variables 
course like Saff and Snider.

Kip Murray


Raul Miller wrote:
> I have in the past thought that quaternions would be a nice addition to J.
> 
> They are very useful, for example, for expressing rotations of geometric 
> models.
> 
> However, J of course needs its implementation to be generally correct
> and not just correct for some specific application.
> 
> So I have been thinking of how to model J's primitives that would need
> to interpret quaternion values.
> 
> This leads to questions such as: what does x^y mean when x and y are
> quaternions?
> 
> Apparently, this is a hard problem:
> 
>    http://www.zipcon.net/~swhite/docs/math/quaternions/analysis.html
> 
> I have been trying to find a definition of quaternion exponentials
> that that does not do something silly (like assume that quaternion
> multiplication is commutative).  So far, I have not found anything I
> feel I could implement a J model for transcendental functions from.
> Also, x^y for quaternion x and quaternion y would often enough have an
> infinite set of valid results, which complicates things.
> 
> So... does anyone have an good ideas about how to handle
> transcendental functions that work with quaternions?
> 
> Thanks,
> 
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to