I am hoping to get Mathematica to find an analytic form for the $n$th derivative of the following function of two variables: $$f(a,b)=\frac{1}{\left((1-a) b\right)^{-k} \left(1-(1-a) b\right)^{k-n}+\left((1-a) b+a\right)^{-k} \left(1-(1-a) b-a\right)^{k-n}}$$ where $(a,b)\in\mathbb{R}^2$, $(k,n)\in\mathbb{N}^2$ are two constants such that $k\le n$, and $f(a,b)\in\mathbb{R}$.

To be more specific, rather than an "$n$th derivative", I am really talking about the $\ell$th and $m$th derivative with respect to a and b, as shown in the following Mathematica code:

f[a_, b_] := 1/(
  1/(((1 - a) b)^k (1 - (1 - a) b)^(n - k)) + 
   1/(((1 - a) b + a)^k (1 - (1 - a) b - a)^(n - k)));
D[f[a, b], {a, l}, {b, m}]

The above code produces the following result (which I am including as a screenshot because I have not been able to make the output readable here in any other format): enter image description here

As far as I can tell, the work done by Mathematica here is unsatisfactory because it still keeps two derivatives, one over $b$ and then one over $a$ (and I cannot understand why it has done the other "simplifications" since, as far as I can tell, they do not really push the ball forward).

This question was spawned following a comment on this other question. The answer to that question uses a workaround for obtaining the result of the nth derivative of a function of one variable. That said, it is unclear to me if or how the proposed method would be transferable to a function of two variables.

This is related to yet another question in which it turns out that Mathematica is (in its current version) capable of producing a satisfactory answer using the DifferenceRoot function (although it is still not clear to me how a potential DifferenceRoot function, if it were to appear in my case, would be used). In this post, Maple was also used, and I have tried that route as well, to no avail in my case.

  • 1
    $\begingroup$ f[a_, b_] := 1/(1/(((1 - a) b)^k (1 - (1 - a) b)^(n - k)) + 1/(((1 - a) b + a)^k (1 - (1 - a) b - a)^(n - k))); k = 3; n = 7; l = 5; m = 7; D[f[a, b], {a, l}, {b, m}] result in the output which takes 20.2 MB. Are you really interested in it? $\endgroup$
    – user64494
    Commented Jun 14 at 10:21
  • $\begingroup$ NO hope to compute n-th derivative for general parameters k and n of function f[a,b] , because is too complex. This is one of those questions that will remain unanswered for many years, probably. I may be wrong :P $\endgroup$ Commented Jun 14 at 12:32
  • $\begingroup$ @user64494 thanks for that test, though I do not think that the fact that Mathematica has a large finite series for specific values is related to the size or existence of the closed form, e.g. Sum[Binomial[n, k] x^k, {k, 0, [Infinity]}] vs Sum[Binomial[n, k] x^k, {k, 0, 17}]. I could be wrong, but I have seen simplifications (for series in particular) not immediately visible to MMA. $\endgroup$
    – tyogi
    Commented Jun 14 at 15:35
  • $\begingroup$ @MariuszIwaniuk you might be right, but without asking I would just be stuck on this problem on my own :) thanks for your feedback, I was hoping your workaround would be applicable in this case. $\endgroup$
    – tyogi
    Commented Jun 14 at 15:38


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.