3
$\begingroup$

I semi-regularly run into a problem when trying to use Norm[] or Abs[] to get a symbolic result with FullSimplify[], Simplify[], etc., where MMA and I don't have the same thing in mind for a useful final result. This basic problem has already been tackled in questions like this: Advice for Mathematica as Mathematician's Aid.

With the default options to say, FullSimplify[], we have the following:

$Assumptions = Element[a | b | c | d, Reals];
FullSimplify[Norm[#]^2 & /@ {a + I b, Sqrt[a + I b], Exp[a + I b]}]

(* {a^2 + b^2, Norm[Sqrt[a + I b]]^2, Norm[E^(a + I b)]^2} *)

To get the result I'm looking for, it is pretty straightforward to add a ComplexityFunction to penalize results I don't want from FullSimplify[], and then add additional TransformationFunctions to get the results I do want, e.g.:

complexity [e_] := 100 Count[e, _Norm, {0, \[Infinity]}] + LeafCount[e];
transformations = {Automatic, # /. Norm[Sqrt[x_]]^2 :> Sqrt[Re[x]^2 + Im[x]^2] &, # /. Norm[Exp[x_]]^2 :> Exp[2 Re[x]] &};
FullSimplify[{a + I b, Sqrt[a + I b], Exp[a + I b]}, ComplexityFunction -> complexity, TransformationFunctions -> transformations] 

(*{a^2 + b^2, Sqrt[a^2 + b^2], E^(2 a)}*)

This works very well for simple cases like this, but it has a few issues. It doesn't generalize very well to arbitrary expressions, e.g. we would need another transformation function to deal with the case of something of the form Norm[Times[a + I b, c + I d]]^2 if the expression we were simplifying had complex numbers being multiplied. It also can run prohibitively slow for a sufficiently complicated expression if too many transformation functions are needed.

Question: Is there any way in the specific case of Norm[] or Abs[] when used with one of the simplifying functions (e.g. FullSimplify[]) to efficiently tell MMA to ignore details about multivalued complex expressions and similar that prevent the simplifications I typically want? Also, is there a more general way of using transformations other than the defaults in a way that is less computationally expensive than pattern matching as above?

$\endgroup$
4
$\begingroup$

Just use ComplexExpand before FullSimplify

Assuming[{Element[{a, b}, Reals]},
 Norm[#]^2 & /@ {a + I b, Sqrt[a + I b], Exp[a + I b]} // ComplexExpand // 
  FullSimplify]

(*  {a^2 + b^2, Sqrt[a^2 + b^2], E^(2 a)}  *)
$\endgroup$
  • $\begingroup$ Right, I forgot about ComplexExpand. I thought I had tried that in the past with ugly complex expressions and got back long messes full of Arg and Conjugate and so on, but I just tried it on a long expression and it gave a good result quickly. Thanks! $\endgroup$ – Kevin Lyons Oct 8 '16 at 3:18
  • $\begingroup$ Also I think ComplexExpand lets us get rid of Assuming since it already assumes everything is real by default. $\endgroup$ – Kevin Lyons Oct 8 '16 at 3:19
  • $\begingroup$ @KevinL - If ComplexExpand gives a form that you don't like, use TargetFunctions option. In general -- and in this case -- you need the Assuming for the FullSimplify $\endgroup$ – Bob Hanlon Oct 8 '16 at 3:21
  • $\begingroup$ I had never noticed TargetFunctions before, it looks really useful for this sort of thing. Thanks again, this is something I will use all the time (and have been doing awkwardly for years now). $\endgroup$ – Kevin Lyons Oct 8 '16 at 5:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.