# Extrema and convergence of sequence

• MHB
• mathmari
In summary, the conversation discusses the behavior of the function $f_n(x)=\frac{x+2n}{x^2+n}$ for $n\in \mathbb{N}$, including the determination of local and global extrema, the behavior for $|x|\rightarrow \infty$, and the existence of certain points $x_1<x_2<x_3<x_4$ where $f_n''(x_i)\cdot f_n''(x_{i+1})<0$. The conversation also covers the pointwise and uniform convergence of the sequence $(f_n)_{n\in \mathbb{N}}$ to a function $f$. The summary concludes with a question about whether the local

#### mathmari

Gold Member
MHB
Hey! :giggle:

For $n\in \mathbb{N}$ let $f_n:\mathbb{R}\rightarrow \mathbb{R}$ given by $f_n(x)=\frac{x+2n}{x^2+n}$.
(a) Determine all (local and global) extrema of $f_n$ and the behaviour for $|x|\rightarrow \infty$. Make a sketch for $f_n$ and $f_n'$. Show that there exists $x_1<x_2<x_3<x_4$ with $f_n''(x_i)\cdot f_n''(x_{i+1})<0$ for $i=1,2,3$. The $x_i$ depend on $n$, they don't have to be given explicitely.
(b) Is the sequence $(f_n)_{n\in \mathbb{N}}$ pointwise or uniformly convergent to a function $f$ ?I have done the following :

(a) The derivative as for $x$ is \begin{align*}f_n'(x)&=\frac{(x+2n)'\cdot (x^2+n)-(x+2n)\cdot (x^2+n)'}{(x^2+n)^2}=\frac{(x^2+n)-(x+2n)\cdot 2x}{(x^2+n)^2}=\frac{x^2+n-2x^2-4nx}{(x^2+n)^2}\\ & =\frac{-x^2-4nx+n}{(x^2+n)^2}\end{align*}
We set that equal to $0$ and we get : \begin{align*}f_n'(x)=0&\Rightarrow \frac{-x^2-4nx+n}{(x^2+n)^2}=0\\ &\Rightarrow -x^2-4nx+n=0\\ &\Rightarrow x_{1,2}=\frac{-(-4n)\pm \sqrt{(-4n)^2-4\cdot (-1)\cdot n}}{2\cdot (-1)}=\frac{4n\pm \sqrt{16n^2+4 n}}{-2}\end{align*}
Outside the roots we have that $f_n'<0$ and so $f_n$ is there decreasing and inside the roots $f_n'>0$ and so $f_n$ is there increasing.

That means that $f_n$ has a minimum at $\frac{4n- \sqrt{16n^2+4 n}}{-2}$ and amaximum at $\frac{4n+ \sqrt{16n^2+4 n}}{-2}$.

Is that correct? Or is it meant tocalculate all extrema (local and global) in an other way? :unsure:If $|x|\rightarrow \infty$ then $f_n$ does to $0$ since the denominator goes faster to infinity, right? :unsure:To show the existence of $x_1<x_2<x_3<x_4$ with $f_n''(x_i)\cdot f_n''(x_{i+1})<0$ do we have to use the mean value theorem? :unsure:
(b) For the pointwise convergence :
\begin{equation*}\lim_{n\rightarrow \infty}f_n(x)=\lim_{n\rightarrow \infty}\frac{x+2n}{x^2+n}=2\end{equation*}
So the sequence converges pointwise to $f(x)=2$.

For the uniform convergence :
\begin{equation*}f_n(x)-f(x)=\frac{x+2n}{x^2+n}-2=\frac{x+2n-2x^2-2n}{x^2+n}=\frac{x-2x^2}{x^2+n}\rightarrow |f_n(x)-f(x)|=\left |\frac{x-2x^2}{x^2+n}\right |\end{equation*} The supremum of that expression doesn't go to zero, since when $x\rightarrow \infty$ then that expression goes to $2$. Therefore $f_n$ doesn't converge uniformly to $2$.

Is that correct and complete? :unsure:

Last edited by a moderator:
mathmari said:
(a)
That means that $f_n$ has a minimum at $\frac{4n- \sqrt{16n^2+4 n}}{-2}$ and a maximum at $\frac{4n+ \sqrt{16n^2+4 n}}{-2}$.

Is that correct? Or is it meant to calculate all extrema (local and global) in an other way?

Correct.
We've found a local minimum and a local maximum.
Are they also global extrema or not? (Wondering)

mathmari said:
If $|x|\rightarrow \infty$ then $f_n$ does to $0$ since the denominator goes faster to infinity, right?

Yep. (Nod)

mathmari said:
To show the existence of $x_1<x_2<x_3<x_4$ with $f_n''(x_i)\cdot f_n''(x_{i+1})<0$ do we have to use the mean value theorem?

I think we should check the zeros of $f_n''$ and see what its signs are between those zeros - just like we did for $f_n'$. (Thinking)
mathmari said:
(b) For the pointwise convergence :
For the uniform convergence :
Is that correct and complete?
All correct and complete. (Sun)

Klaas van Aarsen said:
We've found a local minimum and a local maximum.
Are they also global extrema or not? (Wondering)

Do we have to calculate the values of the function at these points and take into consideration also the behaviour of the function when $|x|\rightarrow \infty$ ?:unsure:
Klaas van Aarsen said:
I think we should check the zeros of $f_n''$ and see what its signs are between those zeros - just like we did for $f_n'$. (Thinking)

The second derivative is \begin{align*}f_n''(x)&=\frac{(-x^2-4nx+n)'\cdot (x^2+n)^2-(-x^2-4nx+n)\cdot ((x^2+n)^2)'}{((x^2+n)^2)^2}\\ & =\frac{(-2x-4n)\cdot (x^2+n)^2-(-x^2-4nx+n)\cdot (2(x^2+n)\cdot 2x)}{(x^2+n)^4}\\ & =\frac{(x^2+n)\cdot \left [(-2x-4n)\cdot (x^2+n)-(-x^2-4nx+n)\cdot (2\cdot 2x)\right ]}{(x^2+n)^4} \\ & =\frac{(x^2+n)\cdot \left [2x^3+12nx^2-6nx-4n^2\right ]}{(x^2+n)^4}\\ & =\frac{2x^3+12nx^2-6nx-4n^2}{(x^2+n)^3}\end{align*}
We set it equal to zero and we get \begin{align*}f_n''(x)=0 & \Rightarrow\frac{2x^3+12nx^2-6nx-4n^2}{(x^2+n)^3}=0 \\ & \Rightarrow 2x^3+12nx^2-6nx-4n^2=0\end{align*}
According to Wolfram the roots are these ones .At the exercise we have that " The $x_i$ depend on $n$, they don't have to be given explicitely. ", so I think they mean that we don't have to calculate the roots, or am I wrong?So it must be also an other way, or not? :unsure:

mathmari said:
Do we have to calculate the values of the function at these points and take into consideration also the behaviour of the function when $|x|\rightarrow \infty$ ?
No need to calculate the function values.
Consider that at minus infinity the function approaches 0. Then it goes down to a local minumum, goes up to a local maximum, and goes down again until at plus infinity it approaches 0.
Can we deduce whether the local extrema are global extrema? (Wondering)
The second derivative is \begin{align*}f_n''(x)&=\frac{(-x^2-4nx+n)'\cdot (x^2+n)^2-(-x^2-4nx+n)\cdot ((x^2+n)^2)'}{((x^2+n)^2)^2}\\ & =\frac{(-2x-4n)\cdot (x^2+n)^2-(-x^2-4nx+n)\cdot (2(x^2+n)\cdot 2x)}{(x^2+n)^4}\\ & =\frac{(x^2+n)\cdot \left [(-2x-4n)\cdot (x^2+n)-(-x^2-4nx+n)\cdot (2\cdot 2x)\right ]}{(x^2+n)^4} \\ & =\frac{(x^2+n)\cdot \left [2x^3+12nx^2-6nx-4n^2\right ]}{(x^2+n)^4}\\ & =\frac{2x^3+12nx^2-6nx-4n^2}{(x^2+n)^3}\end{align*}
We set it equal to zero and we get \begin{align*}f_n''(x)=0 & \Rightarrow\frac{2x^3+12nx^2-6nx-4n^2}{(x^2+n)^3}=0 \\ & \Rightarrow 2x^3+12nx^2-6nx-4n^2=0\end{align*}
According to Wolfram the roots are these ones .At the exercise we have that " The $x_i$ depend on $n$, they don't have to be given explicitely. ", so I think they mean that we don't have to calculate the roots, or am I wrong?So it must be also an other way, or not? :unsure:
It suffices if we can deduce that there are 3 distinct zeros and if we can deduce it switches signs between those zeros.
Consider again the behavior of the function. Where must the inflection points be? (Wondering)

Klaas van Aarsen said:
No need to calculate the function values.
Consider that at minus infinity the function approaches 0. Then it goes down to a local minumum, goes up to a local maximum, and goes down again until at plus infinity it approaches 0.
Can we deduce whether the local extrema are global extrema? (Wondering)

So are the local extrema also global? :unsure:
Klaas van Aarsen said:
It suffices if we can deduce that there are 3 distinct zeros and if we can deduce it switches signs between those zeros.
Consider again the behavior of the function. Where must the inflection points be? (Wondering)

I got stuck right now. Could you explain that further to me? :unsure:

mathmari said:
So are the local extrema also global?

Yes.

mathmari said:
I got stuck right now. Could you explain that further to me?
At minus infinity $f_n$ has a horizontal slope.
Then it decreases to a local minimum.
The slope first became more and more negative, and then became less and less negative until the slope was horizontal in the local minimum.
It means that there is an inflection point between minus infinity and the local minimum.
An inflection point means that $f_n''$ is zero.

Similarly we have an inflection point between the minimum and the maximum.
And we must also have an inflection point between the maximum and plus infinity where we approach a horizontal slope again.

Pick $x_1$ left of the first inflection point where $f_n''(x_1)<0$.
And pick $x_2$ between the first and second inflection point where $f_n''(x_2)>0$.
Then $f_n''(x_1)\cdot f_n''(x_2)<0$.

mathmari said:
To show the existence of $x_1<x_2<x_3<x_4$ with $f_n''(x_i)\cdot f_n''(x_{i+1})<0$ do we have to use the mean value theorem?

We can indeed use the mean value theorem. (Nod)

You have found that there are $a$ and $b$ with $a<b$, $f_n'(a)=0$, and $f_n'(b)=0$.
Since $f_n'$ is differentiable, the mean value theorem tells us that there must be a $c$ with $a<c<b$ for which we have $f_n''(c)=0$.

Can we find if $f_n''$ is zero somewhere for $x<a$? (Wondering)

We can do a function analysis and draw the following chart of zeroes and signs.
\begin{tikzpicture}
\begin{scope}
\draw[latex-latex] (-6,0) node[ left ] {$f_n$} -- (6,0);
\draw foreach \i/\x/\y in {-6/{-\infty}/0, -2.5/{a_n}/{}, 0/{-2n}/0, 2.5/{b_n}/{}, 6/\infty/0} { (\i,0.1) node[above] {$\y$} -- (\i,-0.1) node[below] {$\x$} };
\draw foreach \i/\y in {-4/{descending}, -1.25/{ascending}, 1.25/{ascending}, 4/{descending}} { (\i,0.1) node[above] {$\y$} };
\end{scope}
\begin{scope}[yshift=-2cm]
\draw[latex-latex] (-6,0) node[ left ] {$f_n'$} -- (6,0);
\draw foreach \i/\x/\y in {-6/{-\infty}/0, -2.5/{a_n}/0, 2.5/{b_n}/0, 6/\infty/0} { (\i,0.1) node[above] {$\y$} -- (\i,-0.1) node[below] {$\x$} };
\draw foreach \i/\y in {-4/-, 0/+, 4/-} { (\i,0.1) node[above] {$\y$} };
\end{scope}
\begin{scope}[yshift=-4cm]
\draw[latex-latex] (-6,0) node[ left ] {$f_n''$} -- (6,0);
\draw foreach \i/\x/\y in {-6/{-\infty}/0, -4//0, 0/c/0, 4//0, 6/\infty/0} { (\i,0.1) node[above] {$\y$} -- (\i,-0.1) node[below] {$\x$} };
\draw foreach \i/\x/\y in {-5/{x_1}/-, -2/{x_2}/+, 2/{x_3}/-, 5/{x_4}/+} { (\i,0.1) node[above] {$\y$} (\i,-0.1) node[below] {$\x$} };
\end{scope}
\end{tikzpicture}
It shows where $f_n$, $f_n'$, and $f_n''$ are zero, what their signs are, and how they behave at the infinities.
We can pick $x_i$ as in the diagram can't we?

mathmari said:
For the uniform convergence :
\begin{equation*}f_n(x)-f(x)=\frac{x+2n}{x^2+n}-2=\frac{x+2n-2x^2-2n}{x^2+n}=\frac{x-2x^2}{x^2+n}\rightarrow |f_n(x)-f(x)|=\left |\frac{x-2x^2}{x^2+n}\right |\end{equation*} The supremum of that expression doesn't go to zero, since when $x\rightarrow \infty$ then that expression goes to $2$. Therefore $f_n$ doesn't converge uniformly to $2$.

Is that correct that we check for $\c\rightarrow \infty$ ?
In that way we show that the supremum is greater or equal to $2$ and so the limit when $n\rightarrow \infty$ goes to $2$ or greater, and not to $0$, that's why we conclude that the convergence is not uniform.
Is that correct?

:unsure:

mathmari said:
Is that correct that we check for $\c\rightarrow \infty$ ?
In that way we show that the supremum is greater or equal to $2$ and so the limit when $n\rightarrow \infty$ goes to $2$ or greater, and not to $0$, that's why we conclude that the convergence is not uniform.
Is that correct?
Yep. Correct assuming you meant $x\to\infty$ instead of $\c\rightarrow \infty$. (Nod)

## What is an extremum of a sequence?

An extremum of a sequence is a value that is either the maximum or minimum value in the sequence. It can also refer to the point at which the sequence reaches its maximum or minimum value.

## What is the difference between a local and global extremum?

A local extremum is a value that is greater than or less than its neighboring values, while a global extremum is the maximum or minimum value in the entire sequence.

## What is the convergence of a sequence?

The convergence of a sequence refers to the behavior of the sequence as its terms approach a specific value or limit. A sequence is said to converge if its terms get closer and closer to a specific value as the sequence progresses.

## How is the convergence of a sequence determined?

The convergence of a sequence can be determined by evaluating the limit of the sequence. If the limit exists and is finite, then the sequence is said to converge. If the limit does not exist or is infinite, then the sequence does not converge.

## What is the difference between absolute and conditional convergence?

Absolute convergence refers to a sequence that converges regardless of the order in which its terms are added, while conditional convergence refers to a sequence that only converges when its terms are added in a specific order.