Im not sure if we studied this before or not.
dexp(x) = exp(x) - 1.
exp^[1/2](x) - dexp^[1/2](x) = a_1 x^-1 + a_2 x^-2 + ...
That last equation is conjectured.
A much weaker conjecture is that it holds asymptotically for large x.
( and yes I could involve fake function theory now , but I prefer not , its not the main issue here )
In fact , it is not difficult to show that IF exp^[1/2](x) - dexp^[1/2](x) is a laurent series then the x^0 , x^1 , ... terms are all 0.
This all looks pretty familiar and Im not sure if we considered this before or not.
Since dexp has a parabolic fixpoint , I assume the equation only holds asymptotically for large x.
The reason for that is that if exp^[0.5] is laurent and dexp^[0.5] is not then their difference cannot be laurent.
One could also study exp^[0.5] - 2 sinh^[0.5] but that difference grows to 0 very fast.
exp^[1.5] fast ??
Anyway those differences between half-iterates of similar functions fascinates me.
But in this specific dexp case its quite likely that my fascination comes from stuff I have temporarily forgotten and/or from not understanding dexp^[0.5] well.
So back to dexp^[0.5].
the fixpoint is parabolic.
Now recently I mentioned using fake function theory to get a Taylor anyway.
But again the focus here is not on fake function theory.
What I was intrested in was something like this :
dexp^[0.5] = f ( g(x) ).
where f is a Taylor series. Preferably entire.
And g is an analytic function that is not entire.
Or similar.
For instance
dexp^[0.5](x) = f ( sqrt(x) )
or
dexp^[0.5](x) = f ( ln(x^3 + 1) )
I think that the number of pettals gives A such that
parabolicfixpointfunction^[0.5](x) = f ( x^(1/A) ) or something.
Maybe I need to reconsider parabolic fixpoints again.
Notice the once again returning idea of fake function theory :
parabolic fixpointfunction^[0.5](x) = f ( fake x^(1/A) ).
thereby giving a systematic way of computing fake half-iterates at parabolic fixpoints.
Im probably typing more than thinking and should start thinking and perhaps reading now.
But even if I bump my head in 5 min from now , saying it was trivial , I still want to share this.
regards
tommy1729
dexp(x) = exp(x) - 1.
exp^[1/2](x) - dexp^[1/2](x) = a_1 x^-1 + a_2 x^-2 + ...
That last equation is conjectured.
A much weaker conjecture is that it holds asymptotically for large x.
( and yes I could involve fake function theory now , but I prefer not , its not the main issue here )
In fact , it is not difficult to show that IF exp^[1/2](x) - dexp^[1/2](x) is a laurent series then the x^0 , x^1 , ... terms are all 0.
This all looks pretty familiar and Im not sure if we considered this before or not.
Since dexp has a parabolic fixpoint , I assume the equation only holds asymptotically for large x.
The reason for that is that if exp^[0.5] is laurent and dexp^[0.5] is not then their difference cannot be laurent.
One could also study exp^[0.5] - 2 sinh^[0.5] but that difference grows to 0 very fast.
exp^[1.5] fast ??
Anyway those differences between half-iterates of similar functions fascinates me.
But in this specific dexp case its quite likely that my fascination comes from stuff I have temporarily forgotten and/or from not understanding dexp^[0.5] well.
So back to dexp^[0.5].
the fixpoint is parabolic.
Now recently I mentioned using fake function theory to get a Taylor anyway.
But again the focus here is not on fake function theory.
What I was intrested in was something like this :
dexp^[0.5] = f ( g(x) ).
where f is a Taylor series. Preferably entire.
And g is an analytic function that is not entire.
Or similar.
For instance
dexp^[0.5](x) = f ( sqrt(x) )
or
dexp^[0.5](x) = f ( ln(x^3 + 1) )
I think that the number of pettals gives A such that
parabolicfixpointfunction^[0.5](x) = f ( x^(1/A) ) or something.
Maybe I need to reconsider parabolic fixpoints again.
Notice the once again returning idea of fake function theory :
parabolic fixpointfunction^[0.5](x) = f ( fake x^(1/A) ).
thereby giving a systematic way of computing fake half-iterates at parabolic fixpoints.
Im probably typing more than thinking and should start thinking and perhaps reading now.
But even if I bump my head in 5 min from now , saying it was trivial , I still want to share this.
regards
tommy1729