1

Sequences
SCAN, 2023-2024
Subsequences 2
I Subsequences
1) Denition
Denition 1
Let (un )n∈N be a real sequence. We say that the sequence (vn ) is a
subsequence of (un ) when there exists a function ϕ : N → N strictly
increasing such that for all n ∈ N, vn = uϕ(n) .

In other words, this corresponds to keeping only certain terms of (un ), while

renumbering them (without changing the order).

Example 1
The following sequences are subsequences of (un ) : (u2n ), (u2n+1 ), (un+1 ), (un2 ),
(u3n+7 ) . . .The sequence (u(n−3)2 ) is not a subsequence.

Exercise 1
For all n ∈ N, let un = n2 + 2. Let (vn ) = (u2n+1 ) and (wn ) = (un2 ) :
compute the rst 3 terms of (un ), (vn ), and (wn ).
Subsequences 3

2) Limits of Subsequences
Proposition 1
Let (un ) be a sequence such that lim un = ` ∈ R ∪ {±∞}.
n→+∞
Then any subsequence of (un ) has the limit ` as +∞.
In particular, any subsequence of a convergent sequence is convergent.

Therefore : if a subsequence of (un ) is divergent, or if two subsequences of

(un ) have dierent limits, then (un ) diverges.

Exercise 2
For all n ∈ N, let un = (−1)n arctan(n2 ).
Is the sequence (un ) convergent ?
Subsequences 4

Proposition 2
Let (un ) be a sequence and ` ∈ R ∪ {±∞}.

 lim u2n = `
n→+∞
lim un = ` ⇐⇒ .
n→+∞  lim u2n+1 = `
n→+∞

Exercise 3
(−1)n π 1
 
For all n ∈ N, let un = cos + .
3 n
Study the convergence of the sequence (un ).
Adjacent Sequences 5
II Adjacent Sequences

Denition 2
We say that 2 sequences (un ) and (vn ) are adjacent when one is
increasing, the other is decreasing, and lim (un − vn ) = 0.
n→+∞

Theorem 1
If two sequences are adjacent, then they are convergent and have the
same limit.

Exercise 4
n
1 1
For all n ∈ N∗ , let un = and let vn = un + .
X
k! n!
k=0
Show that (un ) and (vn ) are adjacent.
We then deduce that (un ) converges (and it was shown in the rst year
that thanks to the Taylor-Lagrange formula, n→+∞
lim un = e).
Adjacent Sequences 6
Example 1 (The bisection method)
This method uses adjacent sequences to nd an approximate solution of an
equation.
Let f be a continuous and strictly monotonic function on I = [a, b] such that

f (a) < 0 and f (b) > 0. Then, by the intermediate value theorem, the equation
f (x) = 0 has a unique solution α in I .
We construct by recurrence two sequences (an ) and (bn ) as follows : we set
a0 + b0
a0 = a, b0 = b . Let m1 = .
2
If f (m1 ) ≥ 0, then f changes sign on [a0 , m1 ] and thus α ∈ [a0 , m1 ]. In this case,

we set a1 = a0 and b1 = m1 .
If f (m1 ) < 0, then α ∈ [m1 , b0 ] and we set a1 = m1 and a2 = b1 .

And we repeat the operation on [a1 , b1 ].

We thus construct two adjacent sequences (an ) and (bn ) such that for all n ∈ N,
an ≤ α ≤ bn : they converge to the same limit, which is equal to α.
1
Moreover, bn − an = n
(b − a), so an and bn are approximate values of α
2
1
within
n
(b − a).
2
Recurrence sequences of the form un+1 = f (un ) 7
III Recurrence sequences of the form un+ = f (un )
1

1) Denition
Let f : Df → R be a function, and I be an interval contained in Df .
Denition 3
We say that the interval I is stable under f when f (I ) ⊂ I , i.e. :

∀x ∈ I , f (x) ∈ I .

Proposition 3
Let (un ) be the sequence dened by giving u0 ∈ I and the following
recurrence relation : ∀n ∈ N, un+1 = f (un ).

If I is stable under f , then, for all n ∈ N, un exists and un ∈ I .

Throughout this section, we will now assume that

I is stable under f .
Recurrence sequences of the form un+1 = f (un ) 8

Example 2
u0 = 3
(
Let (un ) be such that √ .
∀n ∈ N, un+1 = 1 + un
√
Let f (x) = 1 + x , with Df = [−1, +∞[.
For example, take I = [0 + ∞[⊂ Df .
Since f is increasing and continuous on I , by the intermediate value
theorem we have : f (I ) = [f (0), +∞
lim f [= [1, +∞[⊂ I . So I is stable under f .
Thus, for all n ∈ N, un exists and un ∈ [0, +∞[.

Exercise 5
What function f corresponds to an arithmetic sequence with common
dierence r ? And a geometric sequence with common ratio q ?
Recurrence sequences of the form un+1 = f (un ) 9

2) Case where f is monotonic

When f is monotonic on I , we can deduce certain results about the
direction of variation of the sequence (un ).
Theorem 3
With the previous notations :
1 if f is increasing on I , then the sequence (un ) is monotonic.

To nd its direction of variation, it suces for example to compare u0

and u1 .
2 if f is decreasing on I , then the extracted sequences (u2n ) and
(u2n+1 ) are monotonic. Furthermore, one is increasing and the other
is decreasing.
To nd their respective directions of variation, it suces for example
to compare u0 and u2 .
Recurrence sequences of the form un+1 = f (un ) 10
Example of sequence with increasing f :

Example of sequence with decreasing f :

Recurrence sequences of the form un+1 = f (un ) 11
Exercise 6
For the sequence (un ) from example 2 : determine its direction of variation,
then deduce that it converges.
Exercise 7
Let (wn ) be such that w0 = 1 and for all n ∈ N, wn+1 = 1 − 2wn3 .
1 Determine the directions of variation of the sequences (w2n ) and

(w2n+1 ).
2 Is it possible for (w2n ) and (w2n+1 ) to converge to the same limit ?

What can be deduced about the sequence (wn ) ?

Exercise 8
(
v ∈R
Let (vn ) be such that 0 .
∀n ∈ N, vn+1 = evn + 3 sin2 (vn )
Determine the directions of variation of (vn ).
Recurrence sequences of the form un+1 = f (un ) 12

Remark : when f is not monotonic, studying the variations of (un ) can

sometimes be very dicult.
For example : let α ∈ R, and (un ) the logistic sequence, dened by :
u0 = 1 and un+1 = αun (1 − un ) for all n ∈ N.
This sequence has a very complicated behavior when α ≥ 3.57.
Recurrence sequences of the form un+1 = f (un ) 13

3) Convergence of the sequence

Denition 4
Let α ∈ I . We say that α is a xed point of f when f (α) = α .

Graphically, this corresponds to the fact that the curve of f intersects the
line y = x at the abscissa α.
Theorem 4
Suppose (un ) converges to ` with ` ∈ I and f is continuous at `.
Then ` is a xed point of f .

Attention : this theorem does not prove that (un ) converges ! It only
allows us to nd the possible limits of (un ) in the case where it converges.
Recurrence sequences of the form un+1 = f (un ) 14

Exercise 9
For the sequence (un ) from example 2 : determine the xed points of f ,
and deduce the limit of (un ).

Exercise 10
For the sequence (vn ) from exercise 8 : show by contradiction that (vn ) is
divergent and determine its limit.

Exercise 11
w0 = 1
(
Let (wn ) be the sequence such that : .
∀n ∈ N, wn+1 = 3wn − 4
Study the variation of (wn ), then its limit.
Recurrence sequences of the form un+1 = f (un ) 15

5) Notions of attractive and repulsive xed points

Denition 5
Let f :I →R such that I is stable under f and let ` ∈ I be a xed point of
f.
1 We say that ` is attractive when there exists a real number r > 0
such that for all u0 ∈ I ∩ [` − r , ` + r ], then any sequence with initial
term u0 such that un+1 = f (un ) converges to `.
2 We say that ` is repulsive when for all u0 ∈ I and any sequence with
initial term u0 such that un+1 = f (un ), either un = ` from a certain
rank, or (un ) does not converge to `.

Remark : there may be xed points that are neither attractive nor
repulsive.
Recurrence sequences of the form un+1 = f (un ) 16

Exercise 12
Let f : x 7→ x 2 .
1 Determine the 2 xed points of f in R.
What can be said about the sequence (un ) when u0 is one of these
xed points ?
2 Let (un ) be a recurrent sequence associated with f with u0 ≥ 0.
Show that for all n ∈ N, un ≥ 0, then that the sequence (un )n≥0 is
monotonic.
3 Study the variation and limit of (un ) when u0 ∈]0, 1[ and when
u0 ∈]1, +∞[.
4 Suppose in this question that u0 < 0.
Using the previous question, study the convergence and limit of (un )
depending on the value of u0 .
5 Deduce that one of the xed points of f is attractive, and the other is
repulsive.
Recurrence sequences of the form un+1 = f (un ) 17

Exercise 13
Let q ∈ R − {1} and let f : x 7→ qx .
Then the only xed point of f is 0.
1 Let (un ) be the sequence dened by u0 ∈ R and un+1 = qun .

What type of sequence is it ?

Study the convergence of (un ) depending on the value of q and that
of u0 .
2 For which values of q is the xed point 0 attractive ? Repulsive ?

Remark (result beyond the scope) : Let ` be a xed point of f .

If f is dierentiable at `, then we can show that if |f 0 (`)| < 1, then ` is
attractive and if |f 0 (`)| > 1, then ` is repulsive.
Newton's Method 18
IV Newton's Method

1) Method idea
This method is due to the mathematicians Isaac Newton (1643-1727) and
Joseph Raphson (circa 1648- circa 1715), and was improved by Thomas
Simpson (1710-1761).
Principle : using recurrent sequences to approximately solve an
equation of the form f (x) = 0.
Let f : [a, b] → R be a dierentiable function and let Cf be its graph on
[a, b].
We assume that there exists a real number α ∈ [a, b] (if possible unique)
such that f (α) = 0.
We want to construct a recurrent sequence (un )n∈N such that
lim un = α.
n→+∞
Thus the terms of (un ) will allow us to obtain approximate values of α.
To construct (un ), we use the tangent method.
Newton's Method 19

We start from a value u0  close  to α such that f 0 (u0 ) 6= 0, and we

consider the tangent T0 to the curve Cf at the point of abscissa u0 .
Then T0 has the equation :
y = f (u0 ) + f 0 (u0 )(x − u0 ).

Since T0 is not horizontal, it intersects the axis (Ox) at a point M1 of

abscissa u1 , and we have :
f (u0 )
u1 = u0 − .
f 0 (u0 )

If the tangent T0 is  close to the curve Cf (i.e., if f is close to its

rst-order Taylor expansion at u0 ), then the obtained value u1 will provide
an approximation of α.
If u1 ∈ [a, b] and f 0 (u1 ) 6= 0, then we repeat the operation starting from u1
to obtain a new value u2 , and so on.
Newton's Method 20
Newton's Method 21

2) Assumptions and construction of the method

Denition 5
Let f : [a, b] −→ R be a function and α ∈ [a, b] such that f (α) = 0.
We assume that f is C 1 class on [a, b] and that f 0 does not vanish on
[a, b].
We then dene the function g on [a, b] as :

f (x)
∀x ∈ [a, b], g (x) = x − .
f 0 (x)

Exercise 14
Let y ∈ [a, b].
Show that y is a xed point of g if and only if f (y ) = 0.
Newton's Method 22

We further assume that [a, b] is stable under g , i.e., :

f (x)
∀x ∈ [a, b], x − ∈ [a, b]
f 0 (x)

Denition 6
We x an element u0 ∈ [a, b] and dene the sequence (un )n∈N by
f (un )
∀n ∈ N, un+1 = g (un ) = un − .
f 0 (un )

Proposition 4
If (un )n∈N converges to a real number `, then f (`) = 0.

Therefore, if α is the unique solution of the equation f (α) = 0 on [a, b]

and if (un ) converges, then its limit is α.
Newton's Method 23
3) Limitations of the method
In most cases, the Newton's method is very eective in nding
approximate values of α.
In particular, if f is convex and strictly monotone on [a, b], we can show
that (un ) converges necessarily when u0 = b.
But be careful with the assumptions :
If f 0 vanishes at certain points in [a, b], then it may happen that at
step n, we have f 0 (un ) = 0, and in this case un+1 does not exist.
If the interval [a, b] is not stable under g , it may happen that at step
n, un is outside the domain of denition of f , and in this case un+1 is
not dened.
There may be cases where the sequence (un ) is well dened, but does
not converge. In this case, for example, we can try to take another
value of u0 .
To compute the terms of the sequence, we need to know the
derivative f 0 explicitly.
Newton's Method 24

4) Convergence rate
Under certain assumptions, when f is C 2 on [a, b], we can show using
Taylor-Lagrange inequality that there exists a constant M > 0 (depending
on the choice of the interval [a, b]) such that :
1
(M|u0 − `|)2 .
n
∀n ∈ N, |un − `| ≤
M
Therefore, if M|u0 − `| < 1, the convergence is very fast : it is said to be
quadratic (the number of exact decimals approximately doubles at each
step).
Newton's Method 25
Example 6 (Approximate solution of the equation x = − ln(x))
Let f : x 7→ x + ln(x).
Exercise : Show that the equation f (x) = 0 has a unique solution α on R
and that α ∈ [e−1 , 1].
1
Newton's method : We choose for example u0 = , with
2
f (un ) un (1 − ln(un ))
un+1 = un − 0 = .
f (un ) un + 1
The rst values of the sequence are :
u0 ≈ 0.500000000000
u1 ≈ 0.564382393520
u2 ≈ 0.567138987715
u3 ≈ 0.567143290399
u4 ≈ 0.567143290410

From the 4th iteration, we have 9 exact decimals, so the convergence is

very fast.