Is this entire expression for tetration legal
#1
I'm going to prove to you an entire expression for tetration in one post for all b greater than eta. Is this valid?

We start by defining:

\( \mathcal{E} f(s) = \frac{1}{\Gamma(-s)} \int_0^\infty f(-u) u^{-s-1}\, \partial u \)

We call this the exponential derivative because:

\( \mathcal{E} f(s) = \frac{d^s f(t)}{dt^s} |_{t=0} \)

Where this complex derivative is evaluated according to the Riemann-Liouville differ-integral where exponentiation is the fix point. By this definition we find:

\( \mathcal{E}(b^s) = \ln(b)^s \)

Now we define a new multiplication across FUNCTIONS; i.e. you cannot plug in numbers into this multiplication. We refer to \( \mathbb{E} \) as the complex linear span of \( e^{\lambda s}\,\,;\,\,\lambda \in \mathbb{C} \).

We now define:

\( \forall \alpha \in \mathbb{C}\,\,\,\,;\,\,\,\,\forall f,g,h \in \mathbb{E} \)

\( f \times g \in \mathbb{E} \)

\( \alpha \times f = \alpha f \)

\( f \times g = g \times f \)

\( (f \times g) \times h = f \times (g \times h) \)

\( f \times (g + h) = (f \times g) + (f \times h) \)

With this we define \( \forall \omega, \nu \in \mathbb{C} \):

\( (\Gamma(\omega + 1) (^\omega e)^s) \times(\Gamma(\nu + 1) (^\nu e)^s) = \Gamma(\omega + \nu + 1) (^{\omega+ \nu } e)^s \)

This result gives us a beautiful isomorphism \( \psi: \mathbb{E} \to \mathbb{C}[s] \):

\( \mathbb{U} = \{ 1, e^s, (^2 e)^s, (^3 e)^s, ..., (^N e)^s,...\} \)

\( \psi(\alpha f + \beta g) = \alpha \psi f + \beta \psi g \)

\( \psi(f \times g) = (\psi f) \cdot (\psi g) \)

\( \psi(\mathcal{E} f) = \frac{d \psi(f)}{ds} \)

\( \psi((^N e)^s) = \frac{s^N}{N!} \)


Now thanks to Newton we have:

\( s^\omega = \sum_{N=0}^{\infty} \frac{\Gamma(\omega + 1)}{\Gamma(\omega - N + 1)N!} (s - 1)^N \)

\( s^{\omega} = \sum_{N=0}^{\infty} \frac{\Gamma(\omega + 1)}{\Gamma(\omega - N + 1)N!} \sum_{k=0}^N \frac{N!}{(N-k)! k!}(-1)^{N-k}s^k \)

Now we apply the isomorphism to get the result \( \exists \phi \in \mathbb{E} \):

\( \Gamma(\omega + 1) \phi_\omega(s) = \sum_{N=0}^{\infty} \frac{\Gamma(\omega + 1)}{\Gamma(\omega - N + 1)N!} \sum_{k=0}^N \frac{N!}{(N-k)! k!}(-1)^{N-k}k!(^k e)^s \)

We can reduce this equation to:

\( \phi_\omega(s) = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!}(^k e)^s}{\Gamma(\omega - N +1)} \)

We know these functions satisfy the equations:

\( \phi_\omega \times \phi_\nu = \frac{\Gamma(\omega + \nu + 1)}{\Gamma(\omega + 1)\Gamma(\nu + 1)} \phi_{\omega + \nu}(s) \)

\( \mathcal{E} \phi_\omega = \phi_{\omega - 1} \)


There is only one function which satisfies this:

\( \phi_{\omega}(s) = (^\omega e)^s \)

We can do the same procedure to arrive at the more general expression:

\( (^\omega b)^s = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!}(^k b)^s}{\Gamma(\omega - N +1)} \)


This is probably the fastest way to derive tetration. It's analytic, entire for \( b > \eta \) when we let \( \Re(s) < 0 \)which is easy to show by the ratio test since tetration grows ridiculously faster than the Gamma function.


Anyone see any mistakes I'm making? Or did I just solve tetration in fifteen minutes of work. LOL

Being quick I'll write the glorious formula; \( \forall b \in \mathbb{R}\,\,;\,\, b > e^{\frac{1}{e}} \):

\( \frac{1}{^\omega b} = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!(^k b)}}{ \Gamma(\omega - N +1)} \)


I'm currently finishing a paper on \( \mathcal{E} \) and this result just happened to fall in my lap. I'm wondering if it's valid so that I can keep it in the paper.
#2
(12/10/2012, 03:43 PM)JmsNxn Wrote: ....Being quick I'll write the glorious formula; \( \forall b \in \mathbb{R}\,\,;\,\, b > e^{\frac{1}{e}} \):

\( \frac{1}{^\omega b} = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!(^k b)}}{ \Gamma(\omega - N +1)} \)
Hey James, very interesting stuff, though its a little over my head. But I did try your formula for base e, for sexp(0.5), \( ^{0.5}e \)
\( \frac{1}{^z e} = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!(^k e)}}{ \Gamma(z - N +1)} \)

I generated the first hundred terms, and took the reciprocal and got the result for sexp(0.5), \( ^{0.5}e \) = -0.0889470119254744974, which is a strange result. Also, the summation also doesn't appear to be converging. Here are the first 100 terms for N=0 to N=99 in the outer loop summation. As you can see, the sum is slowly decreasing (larger negative number), which means the reciprocal is slowly getting smaller, and perhaps approaching zero. By the way, the inner loop converges very nicely with only k=0 to k=3 (four terms) required. For integer values of z, there is also the problem that gamma has a singularity for zero and negative integers, but if you somehow ignore the terms with the singularity, it seems to work.
- Sheldon
Code:
N=0 1.12837916709551257390
N=1 -0.356635834837458935278
N=2 -0.0558854027631410430498
N=3 -0.0206132530877876970179
N=4 -0.0141193897106172441830
N=5 -0.0148207070343604561859
N=6 -0.0178722393015305743419
N=7 -0.0217507480727162365390
N=8 -0.0258849705867867567488
N=9 -0.0300391250051902282142
N=10 -0.0341121984624033791621
N=11 -0.0380621797995234682754
N=12 -0.0418743225766593487278
N=13 -0.0455468128421149043589
N=14 -0.0490839363416516118949
N=15 -0.0524926981203164306902
N=16 -0.0557811147157649591746
N=17 -0.0589573488130329367398
N=18 -0.0620292803860181190169
N=19 -0.0650043080717605755730
N=20 -0.0678892727812415897521
N=21 -0.0706904456867871814295
N=22 -0.0734135490929702590636
N=23 -0.0760637929100568892563
N=24 -0.0786459172599043479163
N=25 -0.0811642360986323262444
N=26 -0.0836226791891616201340
N=27 -0.0860248311369194801384
N=28 -0.0883739669751719546777
N=29 -0.0906730842116677024391
N=30 -0.0929249314732365839175
N=31 -0.0951320339948591280177
N=32 -0.0972967162440782634464
N=33 -0.0994211219794611853738
N=34 -0.101507232030120356963
N=35 -0.103556880061972629931
N=36 -0.105571766571162804793
N=37 -0.107553471319117637288
N=38 -0.109503464398763159097
N=39 -0.111423116098393868509
N=40 -0.113313705708880295616
N=41 -0.115176429401406493529
N=42 -0.117012407286649168777
N=43 -0.118822689752075097934
N=44 -0.120608263161640512378
N=45 -0.122370054991416584391
N=46 -0.124108938465338699673
N=47 -0.125825736747199128971
N=48 -0.127521226738006875002
N=49 -0.129196142521777878991
N=50 -0.130851178497564809986
N=51 -0.132486992230976648134
N=52 -0.134104207054477737117
N=53 -0.135703414442310877300
N=54 -0.137285176182888040280
N=55 -0.138850026368874240602
N=56 -0.140398473222902578280
N=57 -0.141931000774856544457
N=58 -0.143448070404900858692
N=59 -0.144950122264901360038
N=60 -0.146437576589519465340
N=61 -0.147910834907073090058
N=62 -0.149370281159202759044
N=63 -0.150816282737450884553
N=64 -0.152249191444038307074
N=65 -0.153669344383391754809
N=66 -0.155077064790327269347
N=67 -0.156472662800217813649
N=68 -0.157856436165959503295
N=69 -0.159228670926092608054
N=70 -0.160589642028024068590
N=71 -0.161939613909932020797
N=72 -0.163278841044604707844
N=73 -0.164607568448171808999
N=74 -0.165926032156421792429
N=75 -0.167234459671161047302
N=76 -0.168533070378856319747
N=77 -0.169822075943608775267
N=78 -0.171101680676333554542
N=79 -0.172372081881860968129
N=80 -0.173633470185532711615
N=81 -0.174886029840737105782
N=82 -0.176129939018709985571
N=83 -0.177365370081821244524
N=84 -0.178592489841470093563
N=85 -0.179811459801623841895
N=86 -0.181022436388954587949
N=87 -0.182225571170454848279
N=88 -0.183421011059346164175
N=89 -0.184608898510033494726
N=90 -0.185789371702802181506
N=91 -0.186962564718902961836
N=92 -0.188128607706623473485
N=93 -0.189287627038901537828
N=94 -0.190439745462995875203
N=95 -0.191585082242693475689
N=96 -0.192723753293499332602
N=97 -0.193855871311223385361
N=98 -0.194981545894351078562
N=99 -0.196100883660557713855
SUM = -11.2426486101395293427
#3
(12/10/2012, 03:43 PM)JmsNxn Wrote: \( (^\omega b)^s = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!}(^k b)^s}{\Gamma(\omega - N +1)} \)
Mindblowing.

I'll have to look at it a bit more, but this method reminds me a lot of Woon's approach, which many on this forum have lumped in with Newton.

Code:
JmsTetrate[b_, w_, s_, m_] := Sum[Sum[(-1)^(n-k)*Tetrate[b, k]^s/(n-k)!, {k, 0, n}]/Gamma[w - n + 1], {n, 0, m}];

I tested your method with the above snippet, and it works well for small m, but it doesn't take long to blow up and overflow/underflow. I don't remember if I read it on this forum or not, but I believe someone has researched which bases the Newton methods work well for, and perhaps I was using the wrong bases. Anyway, this is certainly an interesting approach, even if it is a spin on something known.

Andrew Robbins

Refs:
S.C.Woon "Analytic Continuation of Operators" http://arxiv.org/abs/hep-th/9707206
#4
(12/10/2012, 03:43 PM)JmsNxn Wrote: Being quick I'll write the glorious formula; \( \forall b \in \mathbb{R}\,\,;\,\, b > e^{\frac{1}{e}} \):

\( \frac{1}{^\omega b} = \sum_{N=0}^{\infty} \frac{\sum_{k=0}^{N} \frac{(-1)^{N-k}}{(N-k)!(^k b)}}{ \Gamma(\omega - N +1)} \)

considering the equation above ; if we do \( exp(-1/x) \) on both sides that should be equivalent to adding 1 to \( \omega \).

If that is not the case then the fundamental functional equation of tetration is not satisfied.

So can you confirm that doing \( exp(-1/x) \) on both sides does what it is suppose to do ?

For those confused notice \( exp(-1/x) = 1/exp(1/x) \) and \( 1/exp^{[n]}(1/x) = (1/exp(1/x))^{[n]}. \)

regards

tommy1729
#5
(12/10/2012, 03:43 PM)JmsNxn Wrote: With this we define \( \forall \omega, \nu \in \mathbb{C} \):

\( (\Gamma(\omega + 1) (^\omega e)^s) \times(\Gamma(\nu + 1) (^\nu e)^s) = \Gamma(\omega + \nu + 1) (^{\omega+ \nu } e)^s \)

I'm lost at this point. Why do you define tetetration to satisfy this equality? How do you know it's true?


Possibly Related Threads…
Thread Author Replies Views Last Post
  Transc entire = exp a + exp b + exp c tommy1729 0 3,629 07/15/2015, 07:37 AM
Last Post: tommy1729
  [entire exp^0.5] The half logaritm. tommy1729 1 5,985 05/11/2014, 06:10 PM
Last Post: tommy1729
  entire function close to sexp ?? tommy1729 8 21,909 04/30/2014, 03:49 PM
Last Post: JmsNxn
  Is the following expression for tetration useful? JmsNxn 9 19,681 11/14/2012, 11:29 PM
Last Post: tommy1729
  Understanding Abel/Schroeder with matrix-expression Gottfried 12 28,374 05/26/2008, 08:45 PM
Last Post: Gottfried



Users browsing this thread: 1 Guest(s)