How to prove that these functions do not intersect?

I want to prove that these two functions $$f(x)$$ and $$g(x)$$ do not intersect for $$x>1$$:
$$f(x)=cosh left(frac{2 sqrt{2} pi x left(x^2-1right) cosh (pi x)}{sqrt{x^4+6 x^2+left(x^2-1right)^2 cosh (2 pi x)+1}}right)$$
$$g(x)=frac{4 x^2+left(x^2-1right)^2 cosh (2 pi x)}{left(x^2+1right)^2}$$
Both functions are greater than one and strictly increasing. Subtraction and taking derivative do not work, the problem becomes more complicated.

Does anyone have an idea to prove it by assuming false assumption? Or to prove that $$f(x)-g(x)$$ has no real root?

Any hints or suggestions are really appreciated.

Mathematics Asked by user805770 on November 21, 2021

Since you said that both functions are greater than one and strictly increasing, they vary so fast that I should instead consider $$F(x)=log(f(x)) qquad text{and} qquad G(x)=log(g(x))$$ If you plot them, you should notice that, as soon as $$x > 3$$, $$F(x)$$ and $$G(x)$$ looks very linear with $$F(x) > G(x)$$; I agree that this does not prove anything.

Now, consider a Taylor expansion of $$F(x)-G(x)$$ around $$x=1$$. This gives $$H(x)=F(x)-G(x)=(x-1)^2 left((pi ^2+1)+left(pi ^2-1right) cosh (2 pi )right)+Oleft((x-1)^3right)$$ So, starting from $$0$$ at $$x=1$$, $$H(x)$$ is an increasing function, which, according to Murphy's principle, will go through a maximum value. This maximum is located very close to $$frac 98$$ and its nature is confirmed by the second derivative test.

For sure, now the problem is : what happens for infinitely large values of $$x$$ ?

For the asymptotics analysis, I shall assume that, for large $$t$$, $$cosh(t) sim frac 12 e^t$$. Factoring the exponentials and ignoring the terms which are divided by $$e^t$$, we end with

$$f(x) sim frac 12 e^{2pi x} quad text{and} quad g(x) sim frac 12 e^{2pi x} frac{left(x^2+1right)^2}{left(x^2-1right)^2}implies H(x) sim frac{4}{x^2}+Oleft(frac{1}{x^6}right)$$ In other words, when $$x$$ is large $$f(x) sim g(x) , exp left({frac{4}{x^2}}right)$$

Computed exactly for $$x=10$$, the result is $$H(10)=0.0400013$$ !!

Edit

Looking at the last result, I computed as few values $$left( begin{array}{cc} x & x^2, H(x) \ 2 & 4.0859321 \ 3 & 4.0165826 \ 4 & 4.0052206 \ 5 & 4.0021354 \ 6 & 4.0010293 \ 7 & 4.0005555 \ 8 & 4.0003256 \ 9 & 4.0002032 \ 10 & 4.0001333 \ 11 & 4.0000911 \ 12 & 4.0000643 \ 13 & 4.0000467 \ 14 & 4.0000347 \ 15 & 4.0000263 \ 16 & 4.0000203 \ 17 & 4.0000160 \ 18 & 4.0000127 \ 19 & 4.0000102 \ 20 & 4.0000083 end{array} right)$$

Answered by Claude Leibovici on November 21, 2021

Related Questions

$a,b,c,d$ such that $a^n=b^n=c^n=d^n=1$ and $a+b+c+d=1$

0  Asked on January 18, 2021 by lio

Using L’Hopital’s rule to find the limit of this function

1  Asked on January 18, 2021 by gregorystory16

Trigonometric graph analysis – point determination

2  Asked on January 18, 2021 by ct-27-3555

Best representation/model of a 3D object from multiple observations

0  Asked on January 18, 2021

Let$A$ be an open, dense set in $mathbb R^n$. Prove that $A + A = mathbb R^n$

1  Asked on January 18, 2021 by francisco-jos-letterio

(proposed) elegant solution to IMO 2003 P1

0  Asked on January 18, 2021 by mnishaurya

Reverse order of polynomial coefficients of type $left(r-xright)^n$

1  Asked on January 18, 2021 by thinkingeye

Should one include already cemented proofs of related principles in one’s paper?

0  Asked on January 18, 2021 by a-kvle

The Cauchy-Crofton formula on a plane

1  Asked on January 18, 2021 by jalede-jale-uff-ne-jale

Solving a limit for capacity of a transmission system

3  Asked on January 18, 2021 by jeongbyulji

Galois group Abstract algebra

1  Asked on January 18, 2021 by user462999

Does there always exists coefficients $c,dinmathbb{R}$ s.t. $ax^3+bx^2+cx+d$ has three different real roots?

2  Asked on January 17, 2021 by w2s

If there is a “worldly ordinal,” then must there be a worldly cardinal?

2  Asked on January 17, 2021 by jesse-elliott

Cover number and matching number in hypergraphs.

1  Asked on January 17, 2021 by josh-ng

$displaystyleint_C (e^x+cos(x)+2y),dx+(2x-frac{y^2}{3}),dy$ in an ellipse

1  Asked on January 17, 2021 by fabrizio-gambeln

Show that $h_n(x)=x^{1+frac{1}{2n-1}}$ converges uniformly on $[-1, 1]$.

1  Asked on January 17, 2021 by wiza

What does the functional monotone class theorem say and how does it relate to the other monotone class theorem?

0  Asked on January 17, 2021 by alan-simonin

Symmetric matrix and Hermitian matrix, unitarily diagonalizable

2  Asked on January 17, 2021