We have \begin{align} U_{0} &= 1 &&\text{and} & V_{0} &= 2 \\ U_{n+1} &= \frac{U_{n}+V_{n}}{2} &&\text{and} & V_{n+1} &= \sqrt{U_{n+1}V_n} \end{align} How to prove two sequence have a common limit?

I found $(U_n)$ is increasing and $(V_n)$ is decreasing but I don't know how to do with $\lim (V_n) - (U_n)$

Can someone help me, please?

  • $\begingroup$ Do you want $U_{n+1} = \frac{U_n+V_n}2$ instead? $\endgroup$ – Somos Sep 15 '18 at 20:55
  • $\begingroup$ @Somos I thought I should use only that limit $V_n - U_n$ but yeah if it's possible to resolve with it , how it works ? $\endgroup$ – KEVIN DLL Sep 15 '18 at 20:58
  • $\begingroup$ I ask because you never tell us what $\,b\,$ is supposed it be. Is it a real number? $\endgroup$ – Somos Sep 15 '18 at 21:00
  • $\begingroup$ @Somos yes i just edit it , $\endgroup$ – KEVIN DLL Sep 15 '18 at 21:02
  • $\begingroup$ What is an "adjacent" sequence? Do you mean convergent? $\endgroup$ – Somos Sep 15 '18 at 21:09

Since $U_n$ is increasing and $V_n$ is decreasing, all you have to show is that the difference $V_n-U_n$ has limit $0$. But, the limit exists since both sequences are monotone and bounded, and rewriting the equation $U_{n+1}=(U_n+V_n)/2$ to $\,V_n = 2 U_{n+1}-U_n \,$ shows the two limits must be equal.

Given a diameter one circle, its circumference is $\,\pi.\,$ Archimedes calculated the perimeters of inscribed and circumscribed regular polygons to find upper and lower bounds for $\,\pi.\,$ Let $\,a(n)\,$ be the perimeter of the circumscribed regular polygon of $\,n\,$ sides, and $\,b(n)\,$ the perimeter of the inscribed regular polygon of $\,n\,$ sides. We have that $\,a(2n)\,$ is the harmonic mean of $\,a(n)\,$ and $\,b(n)\,$ and $\,b(2n)\,$ is the geometric mean of $\,a(2n)\,$ and $\,b(n).\,$

We can start with triangles where $\,n=3\,$ and find $\,a(3) = 3\sqrt{3}\,$ and $\,b(3) = a(3)/2.\,$ We then keep doubling the number of sides indefinitely. The connection between the two recursions is that $\, U_n = 3\sqrt{3}/a(3\,2^n)\,$ and $\, V_n = 3\sqrt{3}/b(3\,2^n)\,$ since the recursions and initial values for $\, U_n, V_n \,$ come from those for $\,a(n), b(n).\,$ The common limit of $\,a(n), b(n)\,$ is $\,\pi,\,$ thus the common limit of $\, U_n, V_n\,$ is $\,3\sqrt{3}/\pi.\,$

If we start with squares where $\,n=4\,$ we find that $\, a(4) = 4\,$ and $\,b(4) = 2\sqrt{2}.\,$ Now let $\, U_n := 4/a(4\, 2^n)\,$ and $\, V_n := 4/b(4\, 2^n).\,$ Then $\, U_0 = 1, \,\, V_0 = \sqrt{2} \,$ and the common limit is $\, 4/\pi.\,$ The general values are $\, b(n) = n\,\sin(\pi/n),\,$ and $\, a(n) = n\,\tan(\pi/n).$


Given that $$ u_{n+1}=\frac{u_n+v_n}2\quad\text{and}\quad v_{n+1}=\sqrt{\vphantom{1}u_{n+1}v_n}\tag1 $$ it is easy to derive that $$ u_{n+1}^2-v_{n+1}^2=\frac{u_n^2-v_n^2}4\tag2 $$ Thus, induction says that $$ u_n^2-v_n^2=\frac{u_0^2-v_0^2}{4^n}\tag3 $$ Furthermore, $$ u_{n+1}-u_n=\frac{v_n-u_n}2\quad\text{and}\quad v_{n+1}^2-v_n^2=\frac{u_n-v_n}2v_n\tag4 $$ Case $\boldsymbol{u_0\gt v_0}$

Equation $(3)$ says $u_n\gt v_n$, and $(4)$ says that $u_n$ is decreasing and $v_n$ is increasing. Since, $u_n\gt v_n\gt v_0$, $u_n$ is decreasing and bounded below. Since $v_n\lt u_n\lt u_0$, $v_n$ is increasing and bounded above.

Case $\boldsymbol{u_0\lt v_0}$

Equation $(3)$ says $u_n\lt v_n$, and $(4)$ says that $u_n$ is increasing and $v_n$ is decreasing. Since, $u_n\lt v_n\lt v_0$, $u_n$ is increasing and bounded above. Since $v_n\gt u_n\gt u_0$, $v_n$ is decreasing and bounded below.

In either case, both $u_n$ and $v_n$ converge. $(3)$ says that their limits are the same.


If $U_n,V_n\in[1,2]$,

$$U_n+V_n\in[2,4]\implies U_{n+1}\in[1,2]$$


$$U_{n+1}V_n\in[1,4]\implies V_n\in[1,2].$$

Hence both sequences are bounded and both converge.

And if the limits exist, they are equal. Indeed,

$$U=\frac{U+V}2,V=\sqrt{UV}$$ imply $U=V$.

  • $\begingroup$ Do you not need some monotony argument? Maybe I overlooked something, but, wouldn't it be also possible that the series just circle around some values? $\endgroup$ – Imago Sep 17 '18 at 16:16
  • $\begingroup$ @Imago: re-read the post, monotonicity is already granted. $\endgroup$ – Yves Daoust Sep 17 '18 at 16:22
  • $\begingroup$ I take this as a yes. It's needed. $\endgroup$ – Imago Sep 17 '18 at 16:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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