A Few Problems
by djmathman, Sep 26, 2014, 4:11 PM
MOSP Homework 2013 wrote:
Let
be a positive integer. Given
non-overlapping circular discs on a rectangular piece of paper, prove that one can cut the piece of paper into convex polygonal pieces each of which contains exactly one disc.


Solution
Here I will devise an algorithm to construct such a set of convex polygonal pieces. The main idea here is to cut along the radical axes of pairwise adjacent circles.
Let
be the
disks on the piece of paper, and mark a point
to denote the radical center of circles
,
, and
.
Pick two circles adjacent to each other, say
and
, on the far left side of the piece of paper. Start cutting along the radical axis of these two circles, and stop when you reach a point of the form
, where
is any integer between
and
not equal to either
or
. Here the radical axes of
and
and of
and
concur; branch off onto these, moving in the direction that ensures you will not cut through a disk. Rinse and repeat. Eventually the piece of paper will be cut into polygonal pieces as desired.
Let






Pick two circles adjacent to each other, say












Remark
This is very hard to rigorize (argh combo is hard), but I really like the idea of radical axes and radical centers. (I've never seen those in a combinatorical geometry problem; I guess there's a first for everything
)

MOSP Homework 2013 wrote:
Let
,
be real numbers such that
. Determine the maximum value of
![\[(a_{n+1}+a_{n+2}+\cdots+a_{2n})-(a_1+a_2+\cdots+a_n).\]](//latex.artofproblemsolving.com/6/d/8/6d86f12525865cbc74c9729105905015ea503d46.png)



![\[(a_{n+1}+a_{n+2}+\cdots+a_{2n})-(a_1+a_2+\cdots+a_n).\]](http://latex.artofproblemsolving.com/6/d/8/6d86f12525865cbc74c9729105905015ea503d46.png)
Solution
Let
be the expression we wish to maximize. Define
to be the sequence of real numbers such that
for all
and
otherwise. The condition can be rewritten as
. Making the appropriate substitutions of variables then gives

As an example, see for yourself that the
case gives
.
Next, we can compute
![\[2\sum_{i=1}^{n-1}i^2+n^2=\dfrac{(n-1)n(2n-1)}3+n^2=\dfrac{n(2n^2+1)}3.\]](//latex.artofproblemsolving.com/f/6/7/f672f703e71f17e49ccf13a1db9f0bf42ca42793.png)
Therefore, by Cauchy-Schwarz,
This is attainable by considering the equality condition of C-S.







As an example, see for yourself that the


Next, we can compute
![\[2\sum_{i=1}^{n-1}i^2+n^2=\dfrac{(n-1)n(2n-1)}3+n^2=\dfrac{n(2n^2+1)}3.\]](http://latex.artofproblemsolving.com/f/6/7/f672f703e71f17e49ccf13a1db9f0bf42ca42793.png)
Therefore, by Cauchy-Schwarz,
![\begin{align*}\left[\sum_{j=1}^{n-1}jb_j+nb_n+\sum_{j=n+1}^{2n-1}(2n-j)b_j\right]^2&\leq\left[\sum_{i=1}^{2n-1}b_i^2\right]\left[\sum_{i=1}^{n-1}i^2+n^2+\sum_{i=n+1}^{2n-1}(2n-i)^2\right]\\&=\dfrac{n(2n^2+1)}3\\\implies S&\leq\boxed{\sqrt{\dfrac{n(2n^2+1)}3}}.\end{align*}](http://latex.artofproblemsolving.com/5/6/5/565021f592b8046cf07e43a80866e602757827f9.png)
Remarks
The problem itself wasn't as difficult as my writeup makes it seem. (Rigorizing/generalizing the summation in the middle is pretty annoying, since it's hard to make ideas such as changing the order of the summation clear.) I was motivated to construct the sequence
during the
case, when Cauchy-Schwarz gave me
![\[\sum_{i=1}^{3}(a_{i+1}-a_i)^2\geq\left[\sum_{i=1}^{3}(a_{i+1}-a_i)\right]^2=(a_{4}-a_1)^2.\]](//latex.artofproblemsolving.com/f/2/5/f2578743ebec65e8f783faff4463fa3cfec9022b.png)
This set a bound for the distances between variables, which led me to define
.


![\[\sum_{i=1}^{3}(a_{i+1}-a_i)^2\geq\left[\sum_{i=1}^{3}(a_{i+1}-a_i)\right]^2=(a_{4}-a_1)^2.\]](http://latex.artofproblemsolving.com/f/2/5/f2578743ebec65e8f783faff4463fa3cfec9022b.png)
This set a bound for the distances between variables, which led me to define

Ultimate Elimination Tourney G2 wrote:
Let
be a point in the interior of triangle
such that
is isosceles and its base angles
and
measures the same as angle
. Reflect
about
and
to get
and
respectively. Suppose that the perpendicular bisector of
intersects
at
. Also, let
intersect
at
and
intersect
at
. Prove that
are collinear.





















Solution
WLOG let
. Let
be the projections of
onto
respectively. Let
be the intersection of
with
and let
and
be the midpoints of
and
respectively. The homothety centered at
with scale factor
sends
to
respectively so it suffices to show
collinear.
Note that
,
, and
are all cyclic quads from the right angles, so from the givens
and
. This in turn implies
. Similarly
.
We now proceed to show
, which immediately implies the result (since
and
are on opposite sides of
). From angle chasing

By considering the reverse configuration on the other side we get
. If these quantities are to be equal, then
, which is true since we're assuming this figure lies in a single plane. Hence
.
Now comes a bit of trig. (I know, I'm sorry D:) To save typing let
be
,
respectively. From Ratio Lemma we get
![\[\dfrac{B_1M}{MC_1}=\dfrac{A_1B_1\sin\angle B_1A_1D}{AC_1\sin\angle DA_1C_1}=\dfrac{A_1B_1\sin\angle DCB_1}{A_1C_1\sin\angle C_1AD}.\]](//latex.artofproblemsolving.com/a/1/7/a175f489c872b18849c58e3332f3de32af3fbd82.png)
Now Law of Sines on
gives
and similarly
. Therefore
![\[\dfrac{B_1M}{MC_1}=\dfrac{A_1B_1}{A_1C_1}\left(\dfrac{DY}{DZ}\right)\left(\dfrac{\sin\angle B}{\sin\angle C}\right)=\dfrac{DY}{DZ}\left(\dfrac{A_1B_1/\sin\angle C}{A_1C_1/\sin\angle B}\right).\]](//latex.artofproblemsolving.com/f/f/8/ff8dca0fa1c69320a614e868fbb6b659b7bb659e.png)
But since
and
have the same circumradius, so do
and
. Therefore the two ratios in the parentheses are equal and the equality reduces to
![\[\dfrac{MB_1}{MC_1}=\dfrac{DY}{DZ}=\dfrac{B_1N}{C_1P}.\]](//latex.artofproblemsolving.com/9/3/7/93725589f6b980441a6c941df6bb510f5418f5ca.png)
Hence
by SAS similarity as desired.
















Note that







We now proceed to show





By considering the reverse configuration on the other side we get



Now comes a bit of trig. (I know, I'm sorry D:) To save typing let



![\[\dfrac{B_1M}{MC_1}=\dfrac{A_1B_1\sin\angle B_1A_1D}{AC_1\sin\angle DA_1C_1}=\dfrac{A_1B_1\sin\angle DCB_1}{A_1C_1\sin\angle C_1AD}.\]](http://latex.artofproblemsolving.com/a/1/7/a175f489c872b18849c58e3332f3de32af3fbd82.png)
Now Law of Sines on



![\[\dfrac{B_1M}{MC_1}=\dfrac{A_1B_1}{A_1C_1}\left(\dfrac{DY}{DZ}\right)\left(\dfrac{\sin\angle B}{\sin\angle C}\right)=\dfrac{DY}{DZ}\left(\dfrac{A_1B_1/\sin\angle C}{A_1C_1/\sin\angle B}\right).\]](http://latex.artofproblemsolving.com/f/f/8/ff8dca0fa1c69320a614e868fbb6b659b7bb659e.png)
But since




![\[\dfrac{MB_1}{MC_1}=\dfrac{DY}{DZ}=\dfrac{B_1N}{C_1P}.\]](http://latex.artofproblemsolving.com/9/3/7/93725589f6b980441a6c941df6bb510f5418f5ca.png)
Hence

Remark
I know, I know, I solved this problem almost a month ago, but I wanted to put this on my blog at some point because it's really nice!
There's really not much motivation I can provide to this solution other than the fact that I was led to prove
through an application of the angle bisector theorem.
There's really not much motivation I can provide to this solution other than the fact that I was led to prove

This post has been edited 5 times. Last edited by djmathman, Mar 5, 2015, 3:43 AM
Reason: align tags
Reason: align tags