Let p(z)=a0+a1z+a2z2+⋯+anzn be a complex
polynomial. Suppose that
1=c0≥c1≥⋯≥cn≥0 is a sequence of real
numbers which is convex (i.e. 2ck≤ck−1+ck+1 for
every k=1,2,…,n−1), and consider the polynomial
q(z)=c0a0+c1a1z+c2a2z2+⋯+cnanzn.
Prove that
∣z∣≤1maxq(z)≤∣z∣≤1maxp(z).
Solution (official)
The polynomials p and q are regular on the complex plane, so by
the Maximum Principle,
max∣z∣≤1∣q(z)∣=max∣z∣=1∣q(z)∣, and similarly for
p. Let us denote Mf=max∣z∣=1∣f(z)∣ for any regular
function f. Thus it suffices to prove that Mq≤Mp.
First, note that we can assume cn=0. Indeed, for cn=1, we
get p=q and the statement is trivial; otherwise,
q(z)=cnp(z)+(1−cn)r(z), where
r(z)=∑j=0n1−cncj−cnajzj. The
sequence cj′=1−cncj−cn also satisfies the
prescribed conditions (it is a positive linear transform of the
sequence cj with c0′=1), but cn′=0 too, so we get
Mr≤Mp. This is enough:
Mq=∣q(z0)∣≤cn∣p(z0)∣+(1−cn)∣r(z0)∣≤cnMp+(1−cn)Mr≤Mp.
Using the Cauchy formulas, we can express the coefficients aj of
p from its values taken over the positively oriented circle
S={∣z∣=1}:
aj=2πi1∫Szj+1p(z)dz=2π1∫Szjp(z)∣dz∣
for 0≤j≤n, otherwise
∫Szjp(z)∣dz∣=0.
Let us use these identities to get a new formula for q, using only
the values of p over S:
2π⋅q(w)=j=0∑ncj(∫Sp(z)z−j∣dz∣)wj.
We can exchange the order of the summation and the integration
(sufficient conditions to do this obviously apply):
2π⋅q(w)=∫S(j=0∑ncj(w/z)j)p(z)∣dz∣.
It would be nice if the integration kernel (the sum between the
brackets) was real. But this is easily arranged — for
−n≤j≤−1, we can add the conjugate expressions, because by
the above remarks, they are zero anyway:
2π⋅q(w)=j=0∑ncj(∫Sp(z)z−j∣dz∣)wj=j=−n∑nc∣j∣(∫Sp(z)z−j∣dz∣)wj,2π⋅q(w)=∫S(j=−n∑nc∣j∣(w/z)j)p(z)∣dz∣=∫SK(w/z)p(z)∣dz∣,
where
K(u)=j=−n∑nc∣j∣uj=c0+2j=1∑ncjℜ(uj)
for u∈S.
Let us examine K(u). It is a real-valued function. Again from the
Cauchy formulas, ∫SK(u)∣du∣=2πc0=2π.
If ∫S∣K(u)∣∣du∣=2π still holds (taking the absolute
value does not increase the integral), then for every w:
2π∣q(w)∣=∫SK(w/z)p(z)∣dz∣≤∫S∣K(w/z)∣⋅∣p(z)∣∣dz∣≤Mp∫S∣K(u)∣∣du∣=2πMp;
this would conclude the proof. So it suffices to prove that
∫S∣K(u)∣∣du∣=∫SK(u)∣du∣, which is to say, K is
non-negative.
Now let us decompose K into a sum using the given conditions for
the numbers cj (including cn=0). Let
dk=ck−1−2ck+ck+1 for k=1,…,n (setting
cn+1=0); we know that dk≥0. Let
Fk(u)=j=−k+1∑k−1(k−∣j∣)uj. Then
K(u)=k=1∑ndkFk(u) by easy induction (or see
Figure for a graphical illustration).
So it suffices to prove that Fk(u)≥0 for u∈S. This is
reasonably well-known (as Fk is the Fejér kernel), and also very
easy:
Fk(u)=(1+u+u2+⋯+uk−1)(1+u−1+u−2+⋯+u−(k−1))==(1+u+u2+⋯+uk−1)(1+u+u2+⋯+uk−1)=1+u+u2+⋯+uk−12≥0
This completes the proof.
How the field did
contestants scored
334
average (of 10)
0.07
solved (≥ 80%)
0.0%
near-0 (≤ 10%)
98.2%
discrimination
0.22
Score distribution (field cohort)
Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.