You are on page 1of 8

n

I recently investigated a strange recursive sequence: xn = xn-1


. I suppose the inspiration came from
the recursive definition for the factorial function, xn = n xn-1 , but it's completely unrelated and so
irrelavant.

So, what does this function look like? We can start out with x0 = 1 just to get the feel for what it
looks like:
2 Recursive Sequence.nb

F
n
In[13]:= x@n_D := NB
x@n - 1D
x@0D = 1
ListPlot@Table@x@nD, 8n, 0, 50, 1<D, Joined ® TrueD

Out[14]= 1

Out[15]=
4

10 20 30 40 50

A few questions arise just from this graph: does it tend to infinity, or near an asymptote? Will chang-
ing the initial value change the function's behavior? Why does it bounce up and down like that?
Since I can't embed dynamic Mathematica files into this website, I'll tell you the answer to the second
one: making x0 negative only flips the graph about the x-axis, and varying it on (0,¥) makes the
"bounces" more dramatic, i.e. increases the distance between the maxes and the mins; this is true as
x0 goes to infinity at least. I also noticed that around x0 = .8 that distance gets very small, and that
observation led me to ask the question: is there a value or range of values for x0 where the function
is monotonic increasing?

To find out, I wrote down what the generic sequence (with initial term x0 = k) and made some
interesting findings. The sequence goes:

k, k1 , 2 k, 3 2*4 k 3*5
2 k , 3 , 2*4 k , ...

Now, we want to know what value of k makes this sequence monotonic increasing. Thus, we get a
number of statements:

k < k1 , 1
k
< 2 k, 2 k < 3
, 3
2k 2k
< 2*4 k
3
, and so on.

How do these help us? They give us an upper bound and a lower bound for k, which we will notice
by writing out two infinite products for these two values. I will use each of those inequalities to
"solve" for k: notice that k will be on the "lesser" side when it is in the numerator (in the even #'d
terms), and on the "greater" side when it is in the denominator (in the odd #'d terms). Thus, we
obtain

1 1* 3 1*3
k < 1, k > , k< 2
, k> , and so on.
2 2* 4
interesting findings. The sequence goes:

Recursive Sequence.nb 3
k, k1 , 2 k, 3 2*4 k 3*5
2 k , 3 , 2*4 k , ...

Now, we want to know what value of k makes this sequence monotonic increasing. Thus, we get a
number of statements:

k < k1 , 1
k
< 2 k, 2 k < 3
, 3
2k 2k
< 2*4 k
3
, and so on.

How do these help us? They give us an upper bound and a lower bound for k, which we will notice
by writing out two infinite products for these two values. I will use each of those inequalities to
"solve" for k: notice that k will be on the "lesser" side when it is in the numerator (in the even #'d
terms), and on the "greater" side when it is in the denominator (in the odd #'d terms). Thus, we
obtain

1 1* 3 1*3
k < 1, k > , k< 2
, k> , and so on.
2 2* 4

After writing out a few more terms for myself, I worked the upper and lower bounds for k that the
two sets of inequalities dictate:

1*3*5... H2 n-1L 2 n+1 1*3*5... H2 n+1L


2*4*6... H2 nL 2*4*6... H2 nL
k< and k > .
2 n+2

Do these two bounds approach the same limit as n ® ¥, or do they approach different values and
create range of values k can take to make this sequence monotonic increasing? (Remember that n
must go to ¥, since the sequence is infinite and thus we have an infinite number of inequalities.)
They can be nested in to products:

2 n + 1 Ûni=1 Ûni=1
2 i-1 1 2 i+1
2i
>k> 2i
2 n+2

But, I have no idea how to find the value of these numerically- they are simple enough that I'm sure
someone has calculated them before, but I'll use mathematica to do the trick for me.

lowerbound@n_D := 1  Sqrt@2 n + 2D ä H2 i + 1L  H2 iL
n
In[22]:=
i=1

upperbound@n_D := Sqrt@2 n + 1D ä H2 i - 1L  H2 iL
n

i=1

Limit@8lowerbound@nD, upperbound@nD<, n ® InfinityD

: >
2 2
Out[24]= ,
Π Π

2
So, it turns out that both of these bounds approach the same value: Π
. We may have been able

to guess that the Π was in there, but the 2 and the square root were unexpected. Below I also cre-
ated a graph to illustrate this asymptote: the x-axis has the number of inequalities we are approximat-
ing to, while the y-axis has the value of k both bounds are giving us.
4 Recursive Sequence.nb

2
So, it turns out that both of these bounds approach the same value: Π
. We may have been able

to guess that the Π was in there, but the 2 and the square root were unexpected. Below I also cre-
ated a graph to illustrate this asymptote: the x-axis has the number of inequalities we are approximat-
ing to, while the y-axis has the value of k both bounds are giving us.

ShowBListPlotBTableB1  Sqrt@2 n + 2D ä H2 i + 1L  H2 iL, 8n, 1, 100<F,


n
In[27]:=
i=1

Joined ® True, PlotStyle ® OrangeF,

ListPlotBTableBSqrt@2 n + 1D ä H2 i - 1L  H2 iL, 8n, 1, 100<F, Joined ® TrueF,


n

i=1

PlotRange ® 8.79, .81<F

0.810

0.805

Out[27]= 0.800

0.795

0 20 40 60 80 100

Before doing this analysis, I tried to algebraically find out if the upper bound would always be
larger than the lower one, and how it would look in the limit (if they would be equal in the limit).
So, I wrote

2 n + 1 Ûni=1 Ûni=1
2 i-1 1 2 i+1
2i
> 2i
2 n+2

Ûni=1 = Ûni=1
2 i+1
2 i+1 3
4 n2 + 6 n + 2 > 2i
2 i-1 2 i-1
= 1
* 53 * 75 ... 2 n+1
2 n-1
= 2n+1
2i

4 n2 + 6 n + 2 > 4 n2 + 4 n + 1
2n+1 > 0

Which is clearly true; eager for results in interpreted this finding as proof that there was a range of
values- however, clearly it only shows that the LHS really is an upper bound and the RHS a lower
bound; in the limit the two are equal.

The next step was to find what xn is explicitly: with a little analysis of the same series written above,
IH-1Li M H-1L
Ixo Ûi=1 i M
n
n
I found the complicated explicit definition: xn = , which isn't particularly interest-
ing or helpful for the mathematics, though it did allow me to overcome a programming issue and
create a dynamic plot of xn that allows x0 to vary. Below is a table of a few of the graphs, including
2
the one with x0 = Π
.
2 n + 1 Ûni=1 Ûni=1
2 i-1 1 2 i+1
2i
> 2i
2 n+2

Ûni=1 = Ûni=1
2 i+1
2 i+1 3
4 n2 + 6 n + 2 > 2i
2 i-1 2 i-1
= 1
* 53 * 75 ... 2 n+1
2 n-1
= 2n+1 Recursive Sequence.nb 5

2i

4 n2 + 6 n + 2 > 4 n2 + 4 n + 1
2n+1 > 0

Which is clearly true; eager for results in interpreted this finding as proof that there was a range of
values- however, clearly it only shows that the LHS really is an upper bound and the RHS a lower
bound; in the limit the two are equal.

The next step was to find what xn is explicitly: with a little analysis of the same series written above,
H-1Ln
I found the complicated explicit definition: xn = Ixo Ûni=1 i IH-1L M M
i
, which isn't particularly interest-
ing or helpful for the mathematics, though it did allow me to overcome a programming issue and
create a dynamic plot of xn that allows x0 to vary. Below is a table of a few of the graphs, including
2
the one with x0 = Π
.

explicit@n_, k_D := k ä Hi ^ HH- 1L ^ iLL ^ HH- 1L ^ nL


n

i=1

Table@ListPlot@Table@explicit@n, kD, 8n, 1, 100<D,


Joined ® True, PlotStyle ® OrangeD, 8k, - 1, 3, .7<D

Out[32]=
6 Recursive Sequence.nb

20 40 60 80 100

-2

-4

Out[32]= : -6 ,

-8

-10

-12

20 40 60 80 100

-5

-10

,
-15

-20

-25

20

15

10 ,

20 40 60 80 100

,
Recursive Sequence.nb 7

14

12

10

8
,
6

20 40 60 80 100

20

15

,
10

20 40 60 80 100

30

25

20

15 >

10

20 40 60 80 100

In[33]:= ListPlot@Table@explicit@n, Sqrt@2  PiDD, 8n, 1, 100<D, Joined ® True, PlotStyle ® OrangeD
8 Recursive Sequence.nb

10

Out[33]=
4

20 40 60 80 100

You might also like