I have a Discord now
by djmathman, Aug 9, 2020, 3:14 AM
oh no
time for my entire social life to disappear?
Russia 1996. A binary operation
on real numbers has the property that
for all
,
,
. Prove that
.
Solution
AoPS. Let
be real numbers. Prove that
![\[
\left(\sum_{k=1}^{n} \frac{a_{k}}{k}\right)^{2} \leq \sum_{k=1}^{n} \sum_{j=1}^{n} \frac{a_{k} a_{j}}{k+j-1}
\]](//latex.artofproblemsolving.com/c/4/2/c427021147b6a4b4df929b6469be046646987031.png)
Quick
Spain 2020 P5. In an acute-angled triangle
, let
be the midpoint of
and
the foot of the altitude to
. Prove that if
, then the circumcircle of triangle
is tangent to
.
Solution![[asy]
size(200);
defaultpen(linewidth(0.7)+fontsize(10));
pair A = origin, B = (2,0);
path El = ellipse((1,0), sqrt(2),1);
pair C = relpoint(El, 0.36);
pair M = (A+B)/2, P = foot(A,C,B);
real r = abs(B-C), s = abs(A-C);
pair T = C * (r+s)/s, N = (A+T)/2, D = s/(r+s)*B;
draw(A--B--C--cycle,rgb(0.1,0.1,0.8));
draw(C--N,rgb(0.1,0.5,0.1));
draw(C--D^^N--M,rgb(0.9,0.1,0.1));
draw(circumcircle(B,M,P),orange+linetype("3 3"));
draw(A--P^^rightanglemark(A,P,C,2),rgb(0.6,0.1,0.6));
clip((-2,-1)--(-2,1.6)--(2.2,1.6)--(2.2,-1)--cycle);
dot("$A$",A,SW);
dot("$B$",B,SE);
dot("$C$",C,NW);
dot("$D$",D,S);
dot("$P$",P,dir(A--P));
dot("$M$",M,S);
dot("$N$",N,NW);
[/asy]](//latex.artofproblemsolving.com/8/2/8/82837c16b10c60f37621b4f091dcc4d3898f00bb.png)
Let
be the point on ray
for which
(not pictured above for space reasons), and let
be the midpoint of
. Then the given length condition implies
or
. This means that the circumcircle of
is tangent to
at
, implying
.
We now claim that
is cyclic. To prove this, let
be the foot of the angle bisector from
. Observe that
is the reflection of
across the external angle bisector of
, so
. Thus, a (simple?) angle chase yields
Thus quadrilateral
is cyclic, as desired.
USOMO 2020 P1, Zuming Feng. Let
be a fixed acute triangle inscribed in a circle
with center
. A variable point
is chosen on minor arc
of
, and segments
and
meet at
. Denote by
and
the circumcenters of triangles
and
, respectively. Determine all points
for which the area of triangle
is minimized.
Solution
CNCM Online R2P7, Albert Wang. A circle is centered at point
in the plane. Distinct pairs of points
and
are diametrically opposite on this circle. Point
is chosen on line segment
such that line
hits the circle again at
and line
at
such that
is the midpoint of
. Now, the point
is taken for
. IF
, find the measure of
.
Solution
Putnam 1985. Evaluate
You may assume that 
Solution
AoPS. Let
be continuous and let
. Consider the sequence
defined by
suppose that the sequence
converges. Show that
has a fixed point.
Solution
Van de Corput Lemma. Let
be
. Suppose that, for some
, we have
for any
, with
monotonic when
. Then
where the constant
is independent of
and
.
Solution (found with hints)
time for my entire social life to disappear?
Russia 1996. A binary operation






Solution
We begin with a lemma.
Lemma. Whenever
,
,
, and
are real numbers such that
,
.
Proof. Observe that
and so
![\[
a*b = ((a*b)*0)*0 = ((c*d)*0)*0 = c*d. \quad\square
\]](//latex.artofproblemsolving.com/b/0/a/b0a315f3cfa5970b94e98b1f7a525a32800bf23a.png)
The above lemma tells us there exists a function
such that
. Plugging this into the original equality gives us
We may furthermore reduce this to the simple two-variable functional equation
. This is now a lot easier to reason with.
Set
to conclude
for all
. Then applying
to both sides of the original equation yields
Now setting
yields
. To determine
, plug back in to the original equality to get
for all
and
; thus
, and so
for all
.
In turn,
for all
and
, and this clearly works.
Lemma. Whenever






Proof. Observe that
![\[
(a*b)*0 = a+b+0 = c+d+0 = (c*d)*0,
\]](http://latex.artofproblemsolving.com/4/8/c/48c3f7e7f1128097429bbdd45f0292eb0a9446e7.png)
![\[
a*b = ((a*b)*0)*0 = ((c*d)*0)*0 = c*d. \quad\square
\]](http://latex.artofproblemsolving.com/b/0/a/b0a315f3cfa5970b94e98b1f7a525a32800bf23a.png)
The above lemma tells us there exists a function


![\[
f(f(a+b)+c) = a+b+c\quad\text{for all }a,b,c\in\mathbb R.
\]](http://latex.artofproblemsolving.com/c/a/d/cadfeea22e8bf2dd7cdfbe508f0a455a462ed8ad.png)

Set




![\[
f(a+b) = f(f(f(a)+b)) = f(a) + b.
\]](http://latex.artofproblemsolving.com/d/3/5/d35fe39183842df2599446f870b05789addab446.png)









In turn,



AoPS. Let

![\[
\left(\sum_{k=1}^{n} \frac{a_{k}}{k}\right)^{2} \leq \sum_{k=1}^{n} \sum_{j=1}^{n} \frac{a_{k} a_{j}}{k+j-1}
\]](http://latex.artofproblemsolving.com/c/4/2/c427021147b6a4b4df929b6469be046646987031.png)
Quick
Let
. Then the Cauchy-Schwarz inequality tells us
![\[
\left(\sum_{k=1}^n\frac{a_k}{k}\right)^2 = \left(\int_0^1 f(x)\,dx\right)^2\leq \int_0^1 f(x)^2\,dx = \sum_{k=1}^n\sum_{j=1}^n\frac{a_ka_j}{k+j-1}.
\]](//latex.artofproblemsolving.com/5/2/4/5240691df03776d115dc41b270d072c73b1f5626.png)

![\[
\left(\sum_{k=1}^n\frac{a_k}{k}\right)^2 = \left(\int_0^1 f(x)\,dx\right)^2\leq \int_0^1 f(x)^2\,dx = \sum_{k=1}^n\sum_{j=1}^n\frac{a_ka_j}{k+j-1}.
\]](http://latex.artofproblemsolving.com/5/2/4/5240691df03776d115dc41b270d072c73b1f5626.png)
Spain 2020 P5. In an acute-angled triangle








Solution
![[asy]
size(200);
defaultpen(linewidth(0.7)+fontsize(10));
pair A = origin, B = (2,0);
path El = ellipse((1,0), sqrt(2),1);
pair C = relpoint(El, 0.36);
pair M = (A+B)/2, P = foot(A,C,B);
real r = abs(B-C), s = abs(A-C);
pair T = C * (r+s)/s, N = (A+T)/2, D = s/(r+s)*B;
draw(A--B--C--cycle,rgb(0.1,0.1,0.8));
draw(C--N,rgb(0.1,0.5,0.1));
draw(C--D^^N--M,rgb(0.9,0.1,0.1));
draw(circumcircle(B,M,P),orange+linetype("3 3"));
draw(A--P^^rightanglemark(A,P,C,2),rgb(0.6,0.1,0.6));
clip((-2,-1)--(-2,1.6)--(2.2,1.6)--(2.2,-1)--cycle);
dot("$A$",A,SW);
dot("$B$",B,SE);
dot("$C$",C,NW);
dot("$D$",D,S);
dot("$P$",P,dir(A--P));
dot("$M$",M,S);
dot("$N$",N,NW);
[/asy]](http://latex.artofproblemsolving.com/8/2/8/82837c16b10c60f37621b4f091dcc4d3898f00bb.png)
Let





![\[
AN = \frac{AC+CB}2 = \frac{AB}{\sqrt 2},
\]](http://latex.artofproblemsolving.com/4/3/c/43cdd9f20d046d60f127129dc4e2e911b9bcc8b8.png)





We now claim that









USOMO 2020 P1, Zuming Feng. Let















Solution
Observe that
so
, and hence by spiral similarity
. Analogously,
.
This has two consequences. First, some angle chasing reveals
Second, explicitly writing out the similarity ratios yields
so
and
are both proportional to
. It follows that
, implying that minimization occurs when
is minimized, i.e.
.
![\[
\angle BO_2D = 2\angle CXB = 2\angle CAB = \angle COB,
\]](http://latex.artofproblemsolving.com/a/1/3/a136dd9fd19c93666c6247e750005427b7bf15ed.png)



![[asy]
//base diagram is from Evan, I just modified things to make my solution work
size(200); defaultpen(linewidth(0.7)); pair C = dir(130); pair A = dir(200); pair B = dir(340); pair X = dir(280); pair D = extension(C, X, A, B); pair O_1 = circumcenter(A, D, X); pair O_2 = circumcenter(B, D, X); pair O = origin; draw(A--O_1--D--O_2--B, orange); filldraw(A--X--B--cycle, invisible, red); filldraw(A--B--C--cycle, invisible, blue); draw(X--C, red); draw(A--O--C^^O--B, darkcyan); draw(unitcircle, grey); draw(O_2--O--O_1,brown);
dot("$C$", C, dir(C)); dot("$A$", A, dir(A)); dot("$B$", B, dir(B)); dot("$X$", X, dir(X)); dot("$D$", D, dir(245)); dot("$O_1$", O_1, S); dot("$O_2$", O_2, S); dot("$O$", O, dir(45));
[/asy]](http://latex.artofproblemsolving.com/6/c/f/6cf6937f61a7f465a99217f3a4a6512719c444e8.png)
![\[
\angle O_1OO_2 = \angle AOB - (\angle AOO_1 + \angle BOO_2) = 2\angle ACB - (\angle ACX + \angle XCB) = \angle ACB.
\]](http://latex.artofproblemsolving.com/5/6/2/5620ccd8a463b55f7287f1affd2898d1784528cd.png)
![\[
\frac{O_2O}{DC} = \frac{BO}{BC}=:\alpha\quad\text{and}\quad \frac{O_1O}{DC} = \frac{AO}{AC}=:\beta,
\]](http://latex.artofproblemsolving.com/b/4/d/b4de7fec4419e2b0b6ae76c21abf4d936eae2303.png)



![$[O_1OO_2]\propto CD^2$](http://latex.artofproblemsolving.com/f/6/5/f65dd95b44ff605650f2a8fb2e5d649754c44aaa.png)


CNCM Online R2P7, Albert Wang. A circle is centered at point















Solution
First observe that
, so
is a 45-45-90 triangle. The fact that
is the midpoint of
implies that
. Combined with
and
, we deduce
is a 45-45-90 triangle as well. Now the
condition implies
In turn, letting
be the center of the circle,
.
Now construct point
so that
is a square.
![[asy]
size(300);
defaultpen(linewidth(0.7)+fontsize(10));
real r = 25;
pair A = (-1,0), B = (1,0), O = origin, D = dir(r), C = -1*D, M = dir(r+90), P = intersectionpoint(M--B,A--D), X = 2*M-P, Q = 2*M - A, Y = reflect(B,Q)*X;
draw(M--O,gray(0.7));
draw(A--B^^C--D,rgb(0.1,0.1,0.7));
draw(unitcircle,rgb(0.1,0.6,0.1));
draw(C--X--B^^A--M,rgb(0.5,0.1,0.5));
draw(A--D,rgb(0.1,0.3,0.1));
draw(X--Y--B,brown);
draw(X--Q--P^^M--Q,red);
draw(B--Q--Y,orange);
dot("$A$",A,W,linewidth(3.3));
dot("$B$",B,E,linewidth(3.3));
dot("$C$",C,dir(O--C),linewidth(3.3));
dot("$D$",D,dir(O--D),linewidth(3.3));
dot("$X$",X,NW,linewidth(3.3));
dot("$Q$",Q,dir(100),linewidth(3.3));
dot("$Y$",Y,dir(70),linewidth(3.3));
dot("$P$",P,S,linewidth(3.3));
dot("$M$",M,dir(O--M));
dot("$O$",O,S);
[/asy]](//latex.artofproblemsolving.com/b/2/2/b224d3076b7d903d9aa9677f5f9c600ab27829fb.png)
Since
is the midpoint of
and
is the midpoint of
,
. This means triangles
and
are similar isosceles triangles. But, actually, they are spirally similar; the dual spiral similarity implies
. The condition
yields that, in fact, these triangles are congruent, so
; in turn,
is the circumcenter of
.
Finally, observe that
-- the former from
, the latter from the spiral similarity -- so
But this when combined with
yields
as well, so
![\[
\angle XCM = \angle MBA = 45^\circ - \angle PAB = \boxed{35^\circ}.
\]](//latex.artofproblemsolving.com/c/0/9/c09bb724b67762609a1f0bf74a028673d3d7260f.png)









![\[
\angle YXB = \angle(BX,CD) = \frac{\widehat{MC}+\widehat{BD}}2 = \frac{\widehat{MD} + \widehat{BD}}2 = \angle MAB.\qquad(*)
\]](http://latex.artofproblemsolving.com/0/5/2/052245c165594d5ce25925b48bd8b07dd25d7e75.png)


Now construct point


![[asy]
size(300);
defaultpen(linewidth(0.7)+fontsize(10));
real r = 25;
pair A = (-1,0), B = (1,0), O = origin, D = dir(r), C = -1*D, M = dir(r+90), P = intersectionpoint(M--B,A--D), X = 2*M-P, Q = 2*M - A, Y = reflect(B,Q)*X;
draw(M--O,gray(0.7));
draw(A--B^^C--D,rgb(0.1,0.1,0.7));
draw(unitcircle,rgb(0.1,0.6,0.1));
draw(C--X--B^^A--M,rgb(0.5,0.1,0.5));
draw(A--D,rgb(0.1,0.3,0.1));
draw(X--Y--B,brown);
draw(X--Q--P^^M--Q,red);
draw(B--Q--Y,orange);
dot("$A$",A,W,linewidth(3.3));
dot("$B$",B,E,linewidth(3.3));
dot("$C$",C,dir(O--C),linewidth(3.3));
dot("$D$",D,dir(O--D),linewidth(3.3));
dot("$X$",X,NW,linewidth(3.3));
dot("$Q$",Q,dir(100),linewidth(3.3));
dot("$Y$",Y,dir(70),linewidth(3.3));
dot("$P$",P,S,linewidth(3.3));
dot("$M$",M,dir(O--M));
dot("$O$",O,S);
[/asy]](http://latex.artofproblemsolving.com/b/2/2/b224d3076b7d903d9aa9677f5f9c600ab27829fb.png)
Since












Finally, observe that


![\[
\angle QXY = \angle XYQ = \angle PYB = 10^\circ.
\]](http://latex.artofproblemsolving.com/4/5/c/45cef796eb94fe0b09fcfb22a40b9b493f7d77bf.png)


![\[
\angle XCM = \angle MBA = 45^\circ - \angle PAB = \boxed{35^\circ}.
\]](http://latex.artofproblemsolving.com/c/0/9/c09bb724b67762609a1f0bf74a028673d3d7260f.png)
Putnam 1985. Evaluate


Solution
Let
. The change of variable
implies
This means
where in
we make the change of variable
. In turn,
.


![\[
I = \int_\infty^0 t^{1/2}e^{-1985\left(t+t^{-1}\right)}\cdot -t^{-2}\,dt = \int_0^\infty t^{-3/2}e^{-1985\left(t+t^{-1}\right)}\,dt.
\]](http://latex.artofproblemsolving.com/8/0/e/80ec2366e6bf623f504df17551c419f06234ceac.png)
![\begin{align*}
2I &= \int_0^\infty (t^{-1/2} + t^{-3/2})e^{-1985\left(t+t^{-1}\right)}\,dt \\
&= e^{-3970}\int_0^\infty (t^{-1/2} + t^{-3/2})e^{-1985\left(t^{1/2}-t^{-1/2}\right)^2}\,dt\\\
&\stackrel{\mathmakebox[\widthof{=}]{(*)}}= 2e^{-3970}\int_0^\infty e^{-1985u^2}\,du = 2e^{-3970}\sqrt{\frac{\pi}{1985}},
\end{align*}](http://latex.artofproblemsolving.com/2/6/8/268fc4b9affbb4c5bc0696cd060266a4c311761e.png)



AoPS. Let






Solution
If
is a fixed point of
then we are done, so assume
; without loss of generality let
. There are two cases to consider.
Case 1:
for some
. Then the Intermediate Value Theorem implies the existence of some
satisfying
.
Case 2:
for all
. A simple inductive argument then tells us
for all
.
Let
. Suppose
for some
. Then
for every
, so
for all such
. Since the former term in the sum tends to zero and the latter term tends to
as
, this average will eventually exceed
, contradicting the fact that the limit of the original sequence is
. Therefore
for all
, and it follows that the sequence
converges to some real number
. This
is then a fixed point for
.
Having exhausted both possible cases, we may conclude.




Case 1:


![$c\in[u_0,x]$](http://latex.artofproblemsolving.com/4/c/5/4c5d00008522a20d07356fffa9bfa49d264dfcb8.png)

Case 2:




Let

















Having exhausted both possible cases, we may conclude.
Van de Corput Lemma. Let




![$x\in[a,b]$](http://latex.artofproblemsolving.com/4/9/4/49432e6cff349a39878727bec9bfcb0efc9fbb7c.png)


![\[
\left|\int_a^be^{i\lambda\phi(x)}\,dx\right|\leq c_k\lambda^{-\frac1k},
\]](http://latex.artofproblemsolving.com/4/8/e/48eb9f73f7de506419135ffb74fac4fe77a1626d.png)



Solution (found with hints)
We proceed by induction on
. Unlike many induction proofs, though, the base case and inductive step are distinct yet equally difficult.
We first show the base case of
. Assume, without loss of generality, that
for any
(we may do this since
is continuous.) Furthermore, since
is monotonic,
does not change sign; we may again assume without loss that
for all
. Write
To bound the first term, use
to write
For the second term, use the Quotient Rule to write
. Our assumption that
is monotonic comes into play here; when pulling the absolute value inside the integral, we may replace
with
. In particular,
which once again can be bounded above by
. Combining both bounds establishes the base case.
For the inductive step, assume the result holds true for some integer
; again, we may assume without loss that
for all
. This implies that
is monotonic, and hence can vanish at most one point in
. Now define the sets
and
via
Observe that the Fundamental Theorem of Calculus (or the Mean Value Theorem) implies
. This means that
is an interval of length at most
. In turn,
is the union of at most two intervals.
Now write
We bound each integral individually.

We first show the base case of


![$x\in[a,b]$](http://latex.artofproblemsolving.com/4/9/4/49432e6cff349a39878727bec9bfcb0efc9fbb7c.png)




![$x\in[a,b]$](http://latex.artofproblemsolving.com/4/9/4/49432e6cff349a39878727bec9bfcb0efc9fbb7c.png)
![\begin{align*}
\int_a^b e^{i\lambda \phi(x)}\,dx &= \int_a^b (e^{i\lambda\phi})'(x) \cdot\frac{1}{i\lambda\phi'(x)}\,dx \\
&= \left[\frac{e^{i\lambda\phi(x)}}{i\lambda\phi'(x)}\right]_a^b - \frac{1}{i\lambda}\int_a^be^{i\lambda\phi(x)}\left(\frac{1}{\phi'}\right)'(x)\,dx.
\end{align*}](http://latex.artofproblemsolving.com/c/6/e/c6ef88fb05e9ee31342b524907c0eff3b43047dd.png)

![\[
\left|\frac{e^{i\lambda\phi(b)}}{i\lambda\phi'(b)} - \frac{e^{i\lambda\phi(a)}}{i\lambda\phi'(a)}\right| \leq \frac{1}{\lambda\phi'(b)} + \frac{1}{\lambda\phi'(a)} \leq \frac{2}{\lambda}.
\]](http://latex.artofproblemsolving.com/6/e/f/6ef2d98b685237aeba8958fcd47099ab05533c67.png)






For the inductive step, assume the result holds true for some integer


![$x\in[a,b]$](http://latex.artofproblemsolving.com/4/9/4/49432e6cff349a39878727bec9bfcb0efc9fbb7c.png)

![$[a,b]$](http://latex.artofproblemsolving.com/8/e/c/8ecbd1ba3da8f2adef66a63f2ab32c47e63fa734.png)


![\begin{align*}
A &= \left\{x\in[a,b]: |\phi^{(k-1)}(x)| < \lambda^{-\frac 1k}\right\},\\
B &= \left\{x\in[a,b]: |\phi^{(k-1)}(x)| \geq \lambda^{-\frac 1k}\right\}.
\end{align*}](http://latex.artofproblemsolving.com/e/2/2/e226953a22a544515567a097f89a6e168cbec01b.png)




Now write
![\[
\int_a^b e^{i\lambda \phi(x)}\,dx = \int_A e^{i\lambda \phi(x)}\,dx + \int_B e^{i\lambda \phi(x)}\,dx.
\]](http://latex.artofproblemsolving.com/4/6/b/46bac19601e20a2f5e299ad7eee034331001bae0.png)
- For
, a simple bound yields
- For
, observe that the function
satisfies
in
, and furthermore
is monotonic. Therefore (implicitly using the fact that
is a union of at most two intervals)