3

How to prove that the following integral assuming $k^2 < 1$ $$\int_{0}^{1} \frac{d x}{\sqrt{1-x^2}\sqrt{1-k^2 x^2}}$$ is convergent?

Assuming[k^2 < 1, 
 Integrate[1/Sqrt[(1 - x^2)*(1 - k^2*x^2)], {x, 0, 1}]]
EllipticK[k^2]
FunctionRange[{EllipticK[k^2], 0 <= k^2 < 1}, k, y, Reals]

y >= Pi/2

It can also be expressed as how to prove convergence of the following elliptic integral?

EllipticK[k^2], (k^2<1)
Artes
  • 57,212
  • 12
  • 157
  • 245
lotus2019
  • 2,091
  • 5
  • 10
  • Integrate[1/Sqrt[(1 - x^2)*(1 - k^2*x^2)], {x, 0, 1}] – cvgmt Jan 31 '22 at 02:20
  • FunctionDomain[EllipticK[k^2], k] – cvgmt Jan 31 '22 at 02:21
  • It is convergent because there is only one singularity on the path, at the endpoint x=1, and it is easuly seen to be integrable by series expansion: `In[188]:= Series[1/Sqrt[(1 - x^2) (1 - k^2 x^2)], {x, 1, 1}, Assumptions -> -1 < k < 1]

    Out[188]= SeriesData[x, 1, { 2^Rational[-1, 2] (1 - k^2)^Rational[-1, 2] (1 - x)^Rational[-1, 2], Rational[-1, 4] 2^Rational[-1, 2] ( 1 - k^2)^Rational[-1, 2] (-1 + k^2)^(-1) (-1 + 5 k^2) ( 1 - x)^Rational[-1, 2]}, 0, 2, 1]`

    – Daniel Lichtblau Jan 31 '22 at 14:08

2 Answers2

8

In fact the original question asks for demonstrating that the following integral is finite, what Matematica can do simply

FullSimplify[ Integrate[1/(Sqrt[1 - x^2] Sqrt[1 - k^2 x^2]), {x, 0, 1}]  
              < Infinity,
             -1 < k < 1 ]

True

A mathematical proof is comparably easy. We can see that for $\; -1< k<1$ and $\;0\leq x <1$ we have $\frac{1}{ \sqrt{1-k^2 x^2}} \leq \frac{1}{\sqrt{1-k^2}}\;$ and so $$\frac{1}{\sqrt{1-x^2} \sqrt{1-k^2 x^2}} \leq \frac{1}{\sqrt{1-k^2}\sqrt{1-x^2}}$$ and since the both sides of this inequality are nonnegative, integration is a generalization of summation we get $$\int_{0}^{1}\frac{dx}{\sqrt{1-x^2} \sqrt{1-k^2 x^2}} \leq \int_{0}^{1}\frac{dx}{\sqrt{1-k^2}\sqrt{1-x^2}} =\frac{\pi}{2\sqrt{1-k^2}}$$ recalling that $\int_{0}^{1}\frac{dx}{\sqrt{1-x^2}}=\frac{\pi}{2}$

Q.E.D.

A counterpart of a crucial step in the proof can be demonstrated simply as well

Simplify[ 1/(Sqrt[1 - x^2] Sqrt[1 - k^2 x^2]) 
          <= 1/(Sqrt[1 - x^2] Sqrt[1 - k^2]), 
                      -1 < k < 1 && 0 <= x < 1]

True

We've got a better upper bound for the complete elliptic integral of the first kind namely $\frac{\pi}{2 \sqrt{1-k^2}}$. Analogously we can find lower bound of the form $\frac{\pi \arcsin(k)}{2 k}$ and in a slightly different way another lower bound $\frac{\log(\frac{1-k}{1+k})}{2 k}$. The first one works for $x$ close to $0$ and the other one for $x$ close to $1$.

We show all the graphs of appropriate functions on the following plot

Plot[{ EllipticK[k^2], Pi/(2 Sqrt[1 - k^2]), Pi/2 (ArcSin[k]/k), 
      Log[(1 + k)/(1 - k)]/(2 k)}, {k, 0, 1}, 
  PlotStyle -> {Thick, Dashed, Dashed, Dashed}, AxesOrigin -> {0, 0}, 
  PlotLegends -> "Expressions", PlotRange -> {0, 5}]

enter image description here

Artes
  • 57,212
  • 12
  • 157
  • 245
1
Integrate[1/Sqrt[(1 - x^2)*(1 - k^2*x^2)], {x, 0, 1}]

ConditionalExpression[EllipticK[k^2], Re[k^2] <= 1 || k^2 ∉ Reals]

FunctionDomain[EllipticK[k^2], k]

-1 < k < 1

The definition domain just indicated that the EllipticK[k^2] converge only in -1 < k < 1 and divergence outside such interval.

cvgmt
  • 72,231
  • 4
  • 75
  • 133
  • That is not so simple in view of FunctionDomain[EllipticK[k^2], k, Complexes] resulting in -1 + k^2 != 0. It's unclear to me whether the integral Integrate[1/Sqrt[(1 - x^2)*(1 - k^2*x^2)], {x, 0, 1}] converges for such values of k. – user64494 Jan 31 '22 at 08:33
  • ClearAll[k]; Integrate[1/Sqrt[(1 - x^2)*(1 - k^2*x^2)], {x, 0, 1}, Assumptions -> k \[Element] Reals && (k^2 > 1 || k^2 < -1)] outputs (-I EllipticK[1 - 1/k^2] + EllipticK[1/k^2])/Abs[k]. – user64494 Jan 31 '22 at 09:59
  • Thank you!@cvgmt@user64494 – lotus2019 Feb 01 '22 at 02:48
  • EllipticK[k^2] does converge for $k>1$ contrary to your statment: "only in -1 < k < 1 and ..." – Artes Feb 24 '23 at 11:46