You are on page 1of 3

# S´election Internationale 2010

Scientiﬁc test : Mathematics
Exercise 1 : A few results on polynomial interpolation
Let f : [a, b] → R be a continuous function. Let n ∈ N, and n + 1 points x
0
< x
1
< < x
n
in
[a, b], all distinct, in increasing order.
1) Show that there is one and only one polynomial p
n
with degree less than n satisfying
p
n
(x
i
) = f(x
i
), i = 0, . . . , n.
2) We assume from now on that f is n + 1 times diﬀerentiable over [a, b]. Show that for any
x ∈ [a, b], there is a point ξ
x
∈ ] min(x, x
i
), max(x, x
i
)[ such that
f(x) −p
n
(x) =
1
(n + 1)!
π
n+1
(x) f
(n+1)

x
),
where π
n+1
(t) :=

n
i=0
(t −x
i
). Deduce from this identity the upper bound
sup
[a,b]
[f −p
n
[ ≤
(b −a)
n+1
(n + 1)!
sup
[a,b]
[f
(n+1)
[.
3) Let R >
b−a
2
, and c :=
a+b
2
. We assume (only for this question) that f is deﬁned over ]c−R, c+R[,
and given there by a power series centered at c:
f(x) =
+∞

k=0
a
k
(x −c)
k
, for all x ∈]c −R, c + R[.
Show that for R large enough, the sequence (p
n
)
n∈N
converges uniformly to f over [a, b].
4) We assume (only for this question) that the points x
i
are uniformly distributed:
x
i
:= a + ih, h :=
b −a
n
.
a) Show that: sup
[a,b]

n+1
[ ≤ h
n+1
max
s∈[0,n]
[s(s −1) (s −n)[
b) Using Stirling’s formula: n! ∼

2πn

n
e

n
, deduce from this inequality: there exists C > 0
(independent of n) such that
sup
[a,b]

n+1
[ ≤ C

b −a
e

n+1
Deduce from this inequlaity an upper bound on sup
[a,b]
[f −p
n
[, better than the one given in 2).
c) Show that there exists c > 0 (independent of n) such that: sup
[a,b]

n+1
[ ≥
c
n

b−a
e

n+1
.
5) We deﬁne, for n ∈ N, t ∈ R, T
n
(t) := cos(narccos t). Show that T
n
is a polynomial with degree
n, satisfying the induction relation
T
n+1
(t) = 2t T
n
(t) − T
n−1
(t), n ∈ N, t ∈ R.
1
Show that T
n
has n roots y
0
< y
1
< < y
n
in [−1, 1], all real and distinct. Determine these roots.
6) We assume (only for this question) that x
i
:=
a+b
2
+
b−a
2
y
i
. Show that sup
[a,b]

n+1
[ =
2

b−a
4

n+1
, and deduce from this identity an upper bound for sup
[a,b]
[f −p
n
[, better than the one
given in 2).
7) Let R
n
[X] the set of all polynomials with real coeﬃcients and degree less than n. Show that
there is p ∈ R
n
[X] such that
sup
[a,b]
[f −p[ = inf
q∈R
n
[X]
sup
[a,b]
[f −q[
Such a polynomial p will be called a polynomial of best approximation of f.
8) We say that g ∈ C([a, b]) equioscillates on (k + 1) points of [a, b] if there are points x
0
< x
1
<
< x
k
in [a, b] such that
For all i = 0, . . . , k, [g(x
i
)[ = |g| and for all i = 0, . . . , k −1, g(x
i+1
) = −g(x
i
).
Show that if p is a polynomial of best approximation , it equioscillates on n + 2 points in [a, b].
9) Deduce from the previous question that there is only one polynomial of best approximation.
10) Show that there is a unique unitary polynomial p with degree n+1 which has minimal uniform
norm over [−1, 1] (|p| = sup
[−1,1]
.[p[). Determine this polynomial.
Exercise 2
Let K be a ﬁeld of characteristic diﬀerent from 2. Recall that a quadratic form (of ﬁnite dimension)
over K is given by a ﬁnite dimensional K-vector space V and a map q : V →K such that
i) q(λv) = λ
2
v for all λ ∈ K and v ∈ V ;
ii) the map q : V V →K deﬁned by (x, y) → q(x, y) =
1
2

q(x + y) −q(x) −q(y)

is bilinear.
Also recall that q → q is a bijection from the set of quadratic forms over K onto the symetric
bilinear forms. In the following, all quadratic forms will be non-degenerate, i.e., for all v ∈ V −¦0¦,
there exists w ∈ V such that q(v, w) ,= 0. An isometry between two quadratic forms (V, q), (V

, q

)
is a bijective linear map f : V

→ V

such that q

◦ f = q.
For all a
1
, . . . , a
n
∈ K − ¦0¦, we denote ¸a
1
, . . . , a
n
¸ the quadratic form deﬁned over K
n
by the
formula q((x
1
, . . . , x
n
)) = a
1
x
2
1
+ + a
n
x
2
n
. Finally we denote Q(K) the set of non-degenerate
Isotropic forms, isometries and orthogonal sum.
We write q

= q

if q and q

are isometric.
1) Show that

= is an equivalence relation on Q(K).
2) A quadratic form q is said to be isotropic if there exists x ,= 0 such that q(x) = 0. Otherwise, q
is said to be anisotropic.
a) Prove that for q ∈ Q(K), there exists x ∈ V such that q(x) ,= 0.
b) Denote H the quadratic form over K
2
deﬁned by H((x
1
, x
2
)) = x
1
x
2
. Show that any isotropic
quadratic form q ∈ Q(K) of dimension 2 is isometric to H and deduce that, for all λ ∈ K − ¦0¦,
one has H

= ¸λ, −λ¸.
2
3) A basis (e
1
, . . . , e
n
) of V is said to be orthogonal for q if q(e
i
, e
j
) = 0 for all i ,= j. Prove that
for any quadratic form q ∈ Q(K), there exists an orthogonal basis (one can use question 2)a) and
consider ¦x¦

= ¦y ∈ V / q(x, y) = 0¦). Deduce that there exists a
1
, . . . , a
n
∈ K − ¦0¦ such that
q

= ¸a
1
, . . . , a
n
¸.
4) Assume q(x) ,= 0. Prove that the map y → s
x
(y) = y − 2
q(x,y)
q(x)
x is an isometry. Deduce that if
q(v) = q(w) ,= 0, then there exists an isometry u from q onto itself such that u(v) = w.
5) Let (V, q), (V

, q

) be two quadratic forms in Q(K). Deﬁne a quadratic form q ⊥ q

: V V

→K
by the formula q ⊥ q

((x, x

)) = q(x) + q

(x

) for all x ∈ V, x

∈ V

.
a) Show that q ⊥ q

is in Q(K).
b) Show that if q
1
, q
2
, q
3
satisfy q
1
⊥ q
3

= q
2
⊥ q
3
, then q
1

= q
2
(one can argue by induction and
use question 4)).
II: Pﬁster forms and similarity factors.
Denote im(q) = ¦a ∈ K−¦0¦ / ∃x ∈ V q(x) = a¦ and Sim(q) = ¦a ∈ K−¦0¦ / aq

= q¦.
1) Prove that Sim(q) is a subgroup of K−¦0¦ that contains all squares of K−¦0¦.
2) Prove that for all quadratic forms q, there exist a unique integer m and an anisotropic quadratic
form q
an
, unique up to an isometry, such that q

= q
an
⊥ mH where mH = H ⊥ ⊥ H. (One can
use Part I).
3) Show that if q is isotropic, im(q) = K−¦0¦.
For all a
1
, . . . , a
n
∈ K−¦0¦ and all quadratic forms (V, q), we denote ¸a
1
, . . . , a
n
form a
1
q ⊥ ⊥ a
n
q over V
n
. A Pﬁster form is a quadratic form of the form
¸¸a
1
, . . . , a
n
¸¸ := ¸1, −a
1
¸ ⊗ ⊗¸1, −a
n
¸.
4) Show that for all a, b, λ ∈ K−¦0¦, one has ¸¸a, b¸¸

= ¸¸−ab, a+b¸¸ and ¸¸a, b¸¸

= ¸¸a, (λ
2
−a)b¸¸.
5) Let p = ¸¸a
1
, . . . , a
n
¸¸ be a Pﬁster form.
a) Show that 1 ∈ im(p) and deduce that there exists φ unique up to an isometry such that
p = ¸1¸ ⊥ −φ.
b) Let b
1
∈ im(φ). Prove that there exists b
2
, . . . , b
n
∈ K − ¦0¦ such that p = ¸¸a
1
, . . . , a
n
¸¸

=
¸¸b
1
, . . . , b
n
¸¸.
c) Let x ∈ V be such that p(x) ,= 0. Show that p

= p(x)p and deduce that Sim(p) = im(p).
III: Application to the level of a ﬁeld.
The level of a ﬁeld K is the least integer n(K) such that −1 is a sum of n(K) squares in K. If such
an integer does not exist, we set n(K) = +∞.
1) Compute the level of the ﬁeld Q of rational numbers, of the ﬁeld C of complex numbers, and of
the ﬁeld Z/pZ for p an odd prime number.
2) Prove that if n(K) is ﬁnite, then n(K) is an (integer) power of 2. (Hint: one can introduce an
adequate Pﬁster form and use Part II).
3