2
$\begingroup$

I'm trying to get a formal power series expansion in Mathematica, e.g. $$\frac{1}{{{x^\alpha } + {x^\beta }}} = \sum\limits_{k = 0}^{ + \infty } {{{( - 1)}^k}{x^{ - \alpha + k(\beta - \alpha )}}} $$ with $\beta > \alpha $, $0 < x < 1$ but Mathematica doesn't want to compute it. Moreover, mathematica doesn't compute it with certain values, e.g.

$Assumptions = 0 < x < 1
Series[1/(x^0.23 + x^0.55), {x, 0, 3}]

So, what should I do to compute these-like "power series"?

Improve this question
$\endgroup$
5
  • 2
    $\begingroup$ try Series[Rationalize[1/(x^0.23 + x^0.55)], {x, 0, 3}]? $\endgroup$
    – kglr
    Commented May 10, 2018 at 6:53
  • $\begingroup$ No, of course, it works with rational powers, but in real, I need to work with any real numbers, so this way does not fit $\endgroup$
    – Afftar43
    Commented May 10, 2018 at 7:33
  • $\begingroup$ @Afftar43 Apply then to the result res proposed by @kglr this: Normal[res] /. Rational[$__] :> N@Rational[$] $\endgroup$
    – Andrew
    Commented May 10, 2018 at 8:19
  • $\begingroup$ For an arbitrary real number you should specify a dx for Rationalize $\endgroup$
    – Bob Hanlon
    Commented May 10, 2018 at 13:50
  • $\begingroup$ (1) A floating point number is actually a rational number (with a power of 2 in the denominator). So those suggesting Rationalize do have a point. (2) Pose something like Series[1/(x^Sqrt[1/2] + x^Sqrt[2/3]), {x, 0, 3}] and folks will see that Rationalize has its limits. $\endgroup$
    – Michael E2
    Commented May 18, 2018 at 18:27

2 Answers 2

Reset to default
1
$\begingroup$

Try factoring and then taking the series. Your expression is equivalent to:

x^(-b)(1/(x^(a - b) + 1))

so

x^(-b) Series[1/(y + 1), {y, 0, 3}]//Normal
(*-(y^3 x^-b) + y^2 x^-b - y x^-b + x^-b*)

(% // Normal) /. y -> x^(a - b)
(*-x^(3*a - 4*b) + x^(2*a - 3*b) - x^(a - 2*b) + x^(-b)*)

The sub anything you want for a and b.

% /. {a -> .22, b -> .55}

(*1/x^1.2100000000000002 - 1/x^0.8800000000000001 - 1/x^1.54 + 1/x^0.55*)
Improve this answer
$\endgroup$
1
$\begingroup$

Let me change and expand the answer of Bill Watts a little bit.

exp = 1/(x^a + x^b)

Numerator[exp] x^-a/(Denominator[exp]  x^-a // Simplify)

(*   x^-a/(1 + x^(-a + b))   *)

Substitute x^(-a + b) -> y and evaluate general series coefficients.

seriescoefficient[k_] = 
      x^-a SeriesCoefficient[1/(y + 1), {y, 0, k}, 
               Assumptions -> k > 0 && Element[k, Integers]]

(*   (-1)^k x^-a   *)

For any analytic function you have

g[y] == Sum[seriescoefficient[k] y^k, {k, 0, Infinity}]

Therefore in this case

seriescoefficient[k] y^k /. y -> x^(-a + b) // PowerExpand

(*   (-1)^k x^(-a + (-a + b) k)   *)

Here the proof

1/(x^a + x^b) == 
     Sum[seriescoefficient[k] y^k /. y -> x^(-a + b), {k, 0, Infinity}]

(*   True   *)
Improve this answer
$\endgroup$

Your Answer

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

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