2
$\begingroup$

I am trying to find the minimum of the function $$f(x)=\int_0^1(1+t^x)\sqrt{1+x^2 t^{2(x-1)}}dt$$ which arises from trying to minimize the surface area of a function rotated around the $x$ axis. Here I am letting that function be equal to $1+t^x$ to obtain an upper bound on the minimum value. Anyway, I used

FindMinimum[ Integrate[(1 + t^x)(1 + x^2 t^(2(x - 1)))^(1/2), {t,0,1}], {x,0,10}]

to find the minimum, and it just kept loading. Is it possible to make it easier for Mathematica to compute it quicker? I have only started to use Mathematica a few days ago and not regularly, so please avoid any "jargon" in your answer.

Edit: After checking Mathematica it says:

enter image description here

This is not true because $f(0)=2$.

Apparently it is lagging this hard?

$\endgroup$
2
  • $\begingroup$ Welcome to the Mathematica Stack Exchange. Does this integral evaluate to a function? $\endgroup$
    – Syed
    Jan 31 at 3:47
  • $\begingroup$ @Syed Yes, notice that we are integrating with respect to $t$ and not $x$. $\endgroup$ Jan 31 at 4:56

1 Answer 1

5
$\begingroup$

Calculating the integral without any assumptions we obtain a ConditionalExpression which may not work quite seamlessly with FindMinimum. When we tackle an involved analytic expression and we expect that it takes a while to find a symbolic integral and a natural remedy might be a numeric definition of the function.

Numeric definition

fn[x_?NumericQ] := NIntegrate[(1 + t^x) (1 + x^2 t^(2 (x - 1)))^(1/2), {t, 0, 1}]
Plot[ fn[x], {x, 0.1, 5}, AxesOrigin -> {0, 2}]

enter image description here

and one can find the minimum, e.g.

FindMinimum[ fn[x], {x, 0.1, 3}]
{2.08127, {x -> 1.65739}}

Symbolic definition

We can also calculate the integral analytically, however it takes a minute or so, but one shouldn't expect much faster processing when hypergeometric functions are involved. We define a piecewise function:

fi[x_, a_] := Integrate[(1 + t^x) (1 + x^2 t^(2 (x - 1)))^(1/2), {t, 0, 1}, 
                Assumptions -> a]
f[x_] = Piecewise @ Table[{fi[x, a], a}, {a, {0 < x < 1, x == 1, x > 1}}];
f[x] // TraditionalForm

enter image description here

and now we can find the maximum e.g.

NMinimize[{f[x], 1 < x < 3}, x]
{2.08127, {x -> 1.65739}} 
Plot[f[x], {x, 0.1, 5}, AxesOrigin -> {0, 2}, 
  Epilog -> {Red, PointSize[0.02], Point[{1.65739, 2.08127}]}]

enter image description here

Sometimes one could also find a symbolic expression for the minimum but this is a different story which cannot be always successful especially when hypergeometric functions are involved.

Surface

RevolutionPlot3D[f[x], {x, 0.01, 3}, Exclusions -> x == 1.65739, 
  Boxed -> False, Axes -> False, 
  PlotStyle -> Directive[Yellow, Specularity[White, 30], Opacity[0.3]],
  RevolutionAxis -> { 1, 0, 0}, Mesh -> {6, 15}, 
  MeshFunctions -> {#1 &, #2 &}, MeshStyle -> {Red, Darker@Cyan}]

enter image description here

$\endgroup$
2
  • $\begingroup$ That’s what I got through Desmos as well. So the moral here is if Mathematica can’t integrate then we make it analytically integrate? $\endgroup$ Jan 31 at 4:57
  • 1
    $\begingroup$ We should take a chance to calculate a numeric integral instead of symbolic one when we expect an immediate result. Analytic expressions are worthy but also expensive. $\endgroup$
    – Artes
    Jan 31 at 5:22

Your Answer

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