Difference between revisions of "2011 USAMO Problems/Problem 3"
(→Solution) |
m (→Solution 2: clarification) |
||
(14 intermediate revisions by 5 users not shown) | |||
Line 2: | Line 2: | ||
==Solutions== | ==Solutions== | ||
===Solution 1=== | ===Solution 1=== | ||
− | Let <math>\angle | + | Let <math>\angle D = \alpha</math>, <math>\angle F = \gamma</math>, and <math>\angle B = \beta</math>, <math>AB=DE=p</math>, <math>BC=EF=q</math>, <math>CD=FA=r</math>. Define the vectors: <cmath>\vec{u} = \vec{AB} + \vec{DE}</cmath> <cmath>\vec{v} = \vec{BC} + \vec{EF}</cmath> <cmath>\vec{w} = \vec{CD} + \vec{FA}</cmath> Clearly, <math>\vec{u}+\vec{v}+\vec{w}=\textbf{0}</math>. |
− | Note that <math>\angle X = 360^\circ - \angle | + | Let <math>AB</math> intersect <math>DE</math> at <math>X</math>. Note that <math>\angle X = 360^\circ - \angle D - \angle C - \angle B = 360^\circ - \alpha - 3\gamma - \beta = 180^\circ - 2\gamma</math>. Define the points <math>M</math> and <math>N</math> on lines <math>AB</math> and <math>DE</math> respectively so that <math>\vec{MX} = \vec{AB}</math> and <math>\vec{XN} = \vec{DE}</math>. Then <math>\vec{u} = \vec{MN}</math>. As <math>XMN</math> is isosceles with <math>XM = XN = p</math>, the base angles are both <math>\gamma</math>. Thus, <math>|\vec{u}|=2p \cos \gamma</math>. Similarly, <math>|\vec{v}|=2q \cos \alpha</math> and <math>|\vec{w}| = 2r \cos \beta</math>. |
− | Next we will find the angles between <math>\vec{u}</math>, <math>\vec{v}</math>, and <math>\vec{w}</math>. As <math>\angle MNX = \gamma</math>, the angle between the vectors <math>\vec{u}</math> and <math>\vec{NE}</math> is <math>\gamma</math>. Similarly, the angle between <math>\vec{ | + | Next we will find the angles between <math>\vec{u}</math>, <math>\vec{v}</math>, and <math>\vec{w}</math>. As <math>\angle MNX = \gamma</math>, the angle between the vectors <math>\vec{u}</math> and <math>\vec{NE}</math> is <math>\gamma</math>. Similarly, the angle between <math>\vec{EF}</math> and <math>\vec{v}</math> is <math>\alpha</math>. Since the angle between <math>\vec{NE}</math> and <math>\vec{EF}</math> is <math>\angle E = 3\beta</math>, the angle between <math>\vec{u}</math> and <math>\vec{v}</math> is <math>360^\circ - \gamma - 3\beta - \alpha = 180^\circ - 2\beta</math>. Similarly, the angle between <math>\vec{v}</math> and <math>\vec{w}</math> is <math>180^\circ - 2\gamma</math>, and the angle between <math>\vec{w}</math> and <math>\vec{u}</math> is <math>180^\circ - 2\alpha</math>. |
− | And since <math>\vec{u}+\vec{v}+\vec{w}=\vec{0}</math>, we can arrange the three vectors to form a triangle, so the triangle with sides of lengths <math>2p \cos \gamma</math>, <math>2q \cos \alpha</math>, and <math>2r \cos \beta</math> has opposite angles of <math>180^\circ - 2\gamma</math>, <math>180^\circ - 2\alpha</math>, and <math>180^\circ - 2\beta</math>, respectively. So by the law of sines: <cmath> \frac{2p \cos \gamma}{\sin 2\gamma} = \frac{2q \cos \alpha}{\sin 2\alpha} = \frac{2r \cos \beta}{\sin 2\beta} </cmath> <cmath> \frac{p}{\sin \gamma} = \frac{q}{\sin \alpha} = \frac{r}{\sin \beta}, </cmath> and the triangle with sides of length <math>p</math>, <math>q</math>, and <math>r</math> has | + | And since <math>\vec{u}+\vec{v}+\vec{w}=\vec{0}</math>, we can arrange the three vectors to form a triangle, so the triangle with sides of lengths <math>2p \cos \gamma</math>, <math>2q \cos \alpha</math>, and <math>2r \cos \beta</math> has opposite angles of <math>180^\circ - 2\gamma</math>, <math>180^\circ - 2\alpha</math>, and <math>180^\circ - 2\beta</math>, respectively. So by the law of sines: <cmath> \frac{2p \cos \gamma}{\sin 2\gamma} = \frac{2q \cos \alpha}{\sin 2\alpha} = \frac{2r \cos \beta}{\sin 2\beta} </cmath> <cmath> \frac{p}{\sin \gamma} = \frac{q}{\sin \alpha} = \frac{r}{\sin \beta}, </cmath> and the triangle with sides of length <math>p</math>, <math>q</math>, and <math>r</math> has corresponding angles of <math>\gamma</math>, <math>\alpha</math>, and <math>\beta</math>. It follows by SAS congruency that this triangle is congruent to <math>FAB</math>, <math>BCD</math>, and <math>DEF</math>, so <math>FD=p</math>, <math>BF=q</math>, and <math>BD=r</math>, and <math>D</math>, <math>F</math>, and <math>B</math> are the reflections of the vertices of triangle <math>ACE</math> about the sides. So <math>AD</math>, <math>BE</math>, and <math>CF</math> concur at the orthocenter of triangle <math>ACE</math>. |
===Solution 2=== | ===Solution 2=== | ||
− | We work in the complex plane, where lowercase letters denote point | + | We work in the complex plane, where lowercase letters denote their corresponding point's poition. Let <math>P</math> denote hexagon <math>ABCDEF</math>. Since <math>AB=DE</math>, the condition <math>AB\not\parallel DE</math> is equivalent to <math>a-b+d-e\ne 0</math>. |
− | Construct a "phantom hexagon" <math>P'=A'B'C'D'E'F'</math> as follows: let <math>A'C'E'</math> be a triangle with <math>\angle{A'C'E'}=\angle{F}</math>, <math>\angle{C'E'A'}=\angle{B}</math>, and <math>\angle{E'A'C'}=\angle{F}</math> (this is possible since <math>\angle{B}+\angle{D}+\angle{F}=180^\circ</math> by the angle conditions), and reflect <math>A',C',E'</math> over its sides to get points <math>D',F',B'</math>, respectively. By rotation and reflection if necessary, we assume <math>A'B'\parallel AB</math> and <math>P',P</math> have the same orientation (clockwise or counterclockwise), i.e. <math>\frac{b-a}{b'-a'}\in\mathbb{R}^+</math>. It's easy to verify that <math>\angle{X'}=\angle{X}</math> for <math>X\in\{A,B,C,D,E,F\}</math> and opposite sides of <math>P'</math> have equal lengths. As the corresponding sides of <math>P</math> and <math>P'</math> must then be parallel, there exist positive reals <math>r,s,t</math> such that <math>r=\frac{a-b}{a'-b'}=\frac{d-e}{d'-e'}</math>, <math>s=\frac{b-c}{b'-c'}=\frac{e-f}{e'-f'}</math>, and <math>t=\frac{c-d}{c'-d'}=\frac{f-a}{f'-a'}</math>. But then <math>0\ne a-b+d-e=r(a'-b'+d'-e')</math>, etc., so | + | Construct a "phantom hexagon" <math>P'=A'B'C'D'E'F'</math> as follows: let <math>A'C'E'</math> be a triangle with <math>\angle{A'C'E'}=\angle{F}</math>, <math>\angle{C'E'A'}=\angle{B}</math>, and <math>\angle{E'A'C'}=\angle{F}</math> (this is possible since <math>\angle{B}+\angle{D}+\angle{F}=180^\circ</math> by the angle conditions), and reflect <math>A',C',E'</math> over its sides to get points <math>D',F',B'</math>, respectively. By rotation and reflection if necessary, we assume <math>A'B'\parallel AB</math> and <math>P',P</math> have the same orientation (clockwise or counterclockwise), i.e. <math>\frac{b-a}{b'-a'}\in\mathbb{R}^+</math>. It's easy to verify that <math>\angle{X'}=\angle{X}</math> for <math>X\in\{A,B,C,D,E,F\}</math> and opposite sides of <math>P'</math> have equal lengths. As the corresponding sides of <math>P</math> and <math>P'</math> must then be parallel, there exist positive reals <math>r,s,t</math> such that <math>r=\frac{a-b}{a'-b'}=\frac{d-e}{d'-e'}</math>, <math>s=\frac{b-c}{b'-c'}=\frac{e-f}{e'-f'}</math>, and <math>t=\frac{c-d}{c'-d'}=\frac{f-a}{f'-a'}</math>. But then <math>0\ne a-b+d-e=r(a'-b'+d'-e')</math>, etc., so the non-parallel condition "transfers" directly from <math>P</math> to <math>P'</math> and |
<cmath>\begin{align*} | <cmath>\begin{align*} | ||
0 | 0 | ||
Line 20: | Line 20: | ||
&=(r-t)(a'-b'+d'-e')+(s-t)(b'-c'+e'-f'). | &=(r-t)(a'-b'+d'-e')+(s-t)(b'-c'+e'-f'). | ||
\end{align*}</cmath> | \end{align*}</cmath> | ||
− | If <math>r-t=s-t=0</math>, then <math>P</math> must be similar to <math>P'</math> and the conclusion is obvious | + | If <math>r-t=s-t=0</math>, then <math>P</math> must be similar to <math>P'</math> and the conclusion is obvious. |
− | Alternatively, use the | + | Otherwise, since <math>a'-b'+d'-e'\ne0</math> and <math>b'-c'+e'-f'\ne0</math>, we must have <math>r-t\ne0</math> and <math>s-t\ne0</math>. Now let <math>x=\frac{a'+d'}{2}</math>, <math>y=\frac{c'+f'}{2}</math>, <math>z=\frac{e'+b'}{2}</math> be the feet of the altitudes in <math>\triangle{A'C'E'}</math>; by the non-parallel condition in <math>P'</math>, <math>x,y,z</math> are pairwise distinct. But <math>\frac{z-x}{z-y}=\frac{s-t}{r-t}\in\mathbb{R}</math>, whence <math>x,y,z</math> are three distinct collinear points, which is clearly impossible. (The points can only be collinear when <math>\triangle{A'C'E'}</math> is a right triangle, but in this case two of <math>x,y,z</math> must coincide.) |
+ | |||
+ | Alternatively (for the previous paragraph), WLOG assume that <math>(A'C'E')</math> is the unit circle, and use the fact that <math>b'=a'+c'-\frac{a'c'}{e'}</math>, etc. to get simple expressions for <math>a'-b'+d'-e'</math> and <math>b'-c'+e'-f'</math>. | ||
+ | |||
+ | ===Solution 3=== | ||
+ | We work in the complex plane to give (essentially) a complete characterization when the parallel condition is relaxed. | ||
+ | |||
+ | WLOG assume <math>a,b,c</math> are on the unit circle. It suffices to show that <math>a,b,c</math> uniquely determine <math>d,e,f</math>, since we know that if we let <math>E</math> be the reflection of <math>B</math> over <math>AC</math>, <math>D</math> be the reflection of <math>A</math> over <math>CE</math>, and <math>F</math> be the reflection of <math>C</math> over <math>AE</math>, then <math>ABCDEF</math> satisfies the problem conditions. (*) | ||
+ | |||
+ | It's easy to see with the given conditions that | ||
+ | <cmath>\begin{align*} | ||
+ | (a-b)(c-d)(e-f) &= (b-c)(d-e)(f-a) \Longleftrightarrow f=\frac{(a-b)(c-d)e+(c-b)(e-d)a}{(a-b)(c-d)+(c-b)(e-d)} \\ | ||
+ | \frac{(e-a)(c-b)}{(a-b)(c-d)+(c-b)(e-d)} = \frac{f-e}{d-e} &= \left(\frac{c-b}{a-b}\right)^2 \overline{\left(\frac{a-b}{c-b}\right)} = \frac{c-b}{a-b}\cdot\frac{c}{a} \Longleftrightarrow d=\frac{c[(a-b)c+(c-b)e]+a(a-e)(a-b)}{c[(a-b)+(c-b)]} \\ | ||
+ | \frac{(a-b)(c-d)+(c-b)(e-d)}{(a-e)(c-d)} = \frac{b-a}{f-a} &= \left(\frac{e-d}{c-d}\right)^2 \overline{\left(\frac{c-d}{e-d}\right)}. | ||
+ | \end{align*}</cmath> | ||
+ | Note that | ||
+ | <cmath>\frac{e-d}{c-d}=\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)},</cmath> | ||
+ | so plugging into the third equation we have | ||
+ | <cmath>\begin{align*} | ||
+ | \frac{a(a-b)(2b-a-c)}{c(c-e)(c-b)-a(a-e)(a-b)} | ||
+ | &=\frac{(a-b)+(c-b)\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)}}{(a-e)}\\ | ||
+ | &=\left(\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)}\right)^2\overline{\left(\frac{c(c-e)(c-b)-a(a-e)(a-b)}{(a-b)[c(e-c)+a(e-a)]}\right)}\\ | ||
+ | &=\left(\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)}\right)^2\frac{\frac{1}{c}\left(\overline{e}-\frac{1}{c}\right)\frac{b-c}{bc}-\frac{1}{a}\left(\overline{e}-\frac{1}{a}\right)\frac{b-a}{ba}}{\frac{b-a}{ab}\left(\frac{1}{c}\left(\frac{1}{c}-\overline{e}\right)+\frac{1}{a}\left(\frac{1}{a}-\overline{e}\right)\right)}\\ | ||
+ | &=\left(\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)}\right)^2\frac{c^3(a\overline{e}-1)(a-b)-a^3(c\overline{e}-1)(c-b)}{c(a-b)[a^2(c\overline{e}-1)+c^2(a\overline{e}-1)]}\\ | ||
+ | &=\frac{(a-b)[c(e-c)+a(e-a)]^2}{[c(c-e)(c-b)-a(a-e)(a-b)]^2}\frac{c^3(a\overline{e}-1)(a-b)-a^3(c\overline{e}-1)(c-b)}{c[a^2(c\overline{e}-1)+c^2(a\overline{e}-1)]}. | ||
+ | \end{align*}</cmath> | ||
+ | Simplifying, this becomes | ||
+ | <cmath>\begin{align*} | ||
+ | &ac(2b-a-c)[c(c-e)(c-b)-a(a-e)(a-b)][a^2(c\overline{e}-1)+c^2(a\overline{e}-1)]\\ | ||
+ | &=[c(e-c)+a(e-a)]^2[c^3(a\overline{e}-1)(a-b)-a^3(c\overline{e}-1)(c-b)]. | ||
+ | \end{align*}</cmath> | ||
+ | Of course, we can also "conjugate" this equation -- a nice way to do this is to note that if | ||
+ | <cmath>x=\frac{(a-b)[c(e-c)+a(e-a)]}{c(c-e)(c-b)-a(a-e)(a-b)},</cmath> | ||
+ | then | ||
+ | <cmath>\frac{a(2b-a-c)}{c(e-c)+a(e-a)}=\frac{x}{\overline{x}}=\overline{\left(\frac{c(e-c)+a(e-a)}{a(2b-a-c)}\right)}=\frac{\frac{1}{c}\left(\overline{e}-\frac{1}{c}\right)+\frac{1}{a}\left(\overline{e}-\frac{1}{a}\right)}{\frac{1}{a}\left(\frac{2}{b}-\frac{1}{a}-\frac{1}{c}\right)},</cmath> | ||
+ | whence | ||
+ | <cmath>ac(2b-a-c)[2ac-b(a+c)]=b[c(e-c)+a(e-a)][a^2(c\overline{e}-1)+c^2(a\overline{e}-1)].</cmath> | ||
+ | If <math>a+c\ne 0</math>, then eliminating <math>\overline{e}</math>, we get | ||
+ | <cmath>e\in\left\{a+c-\frac{ac}{b},a+\frac{2c(c-b)}{a+c},c+\frac{2a(a-b)}{a+c}\right\}.</cmath> | ||
+ | The first case corresponds to (*) (since <math>a,b,c,e</math> uniquely determine <math>d</math> and <math>f</math>), the second corresponds to <math>AB\parallel DE</math> (or equivalently, since <math>AB=DE</math>, <math>a-b=e-d</math>), and by symmetry, the third corresponds to <math>CB\parallel FE</math>. | ||
+ | |||
+ | Otherwise, if <math>c=-a</math>, then we easily find <math>b^2e=a^4\overline{e}</math> from the first of the two equations in <math>e,\overline{e}</math> (we actually don't need this, but it tells us that the locus of working <math>e</math> is a line through the origin). It's easy to compute <math>d=e+\frac{a(a-b)}{b}</math> and <math>f=e+\frac{a(a+b)}{b}</math>, so <math>a-c=2a=f-d\implies c-d=a-f\implies CD\parallel AF</math>, and we're done. | ||
+ | |||
+ | '''Comment.''' It appears that taking <math>(ABC)</math> the unit circle is nicer than, say <math>e=0</math> or <math>(ACE)</math> the unit circle (which may not even be reasonably tractable). | ||
+ | |||
+ | {{MAA Notice}} | ||
==See Also== | ==See Also== | ||
{{USAMO newbox|year=2011|num-b=2|num-a=4}} | {{USAMO newbox|year=2011|num-b=2|num-a=4}} | ||
+ | |||
+ | [[Category:Olympiad Geometry Problems]] | ||
+ | {{MAA Notice}} |
Latest revision as of 11:07, 31 August 2023
In hexagon , which is nonconvex but not self-intersecting, no pair of opposite sides are parallel. The internal angles satisfy
,
, and
. Furthermore
,
, and
. Prove that diagonals
,
, and
are concurrent.
Solutions
Solution 1
Let ,
, and
,
,
,
. Define the vectors:
Clearly,
.
Let intersect
at
. Note that
. Define the points
and
on lines
and
respectively so that
and
. Then
. As
is isosceles with
, the base angles are both
. Thus,
. Similarly,
and
.
Next we will find the angles between ,
, and
. As
, the angle between the vectors
and
is
. Similarly, the angle between
and
is
. Since the angle between
and
is
, the angle between
and
is
. Similarly, the angle between
and
is
, and the angle between
and
is
.
And since , we can arrange the three vectors to form a triangle, so the triangle with sides of lengths
,
, and
has opposite angles of
,
, and
, respectively. So by the law of sines:
and the triangle with sides of length
,
, and
has corresponding angles of
,
, and
. It follows by SAS congruency that this triangle is congruent to
,
, and
, so
,
, and
, and
,
, and
are the reflections of the vertices of triangle
about the sides. So
,
, and
concur at the orthocenter of triangle
.
Solution 2
We work in the complex plane, where lowercase letters denote their corresponding point's poition. Let denote hexagon
. Since
, the condition
is equivalent to
.
Construct a "phantom hexagon" as follows: let
be a triangle with
,
, and
(this is possible since
by the angle conditions), and reflect
over its sides to get points
, respectively. By rotation and reflection if necessary, we assume
and
have the same orientation (clockwise or counterclockwise), i.e.
. It's easy to verify that
for
and opposite sides of
have equal lengths. As the corresponding sides of
and
must then be parallel, there exist positive reals
such that
,
, and
. But then
, etc., so the non-parallel condition "transfers" directly from
to
and
If
, then
must be similar to
and the conclusion is obvious.
Otherwise, since and
, we must have
and
. Now let
,
,
be the feet of the altitudes in
; by the non-parallel condition in
,
are pairwise distinct. But
, whence
are three distinct collinear points, which is clearly impossible. (The points can only be collinear when
is a right triangle, but in this case two of
must coincide.)
Alternatively (for the previous paragraph), WLOG assume that is the unit circle, and use the fact that
, etc. to get simple expressions for
and
.
Solution 3
We work in the complex plane to give (essentially) a complete characterization when the parallel condition is relaxed.
WLOG assume are on the unit circle. It suffices to show that
uniquely determine
, since we know that if we let
be the reflection of
over
,
be the reflection of
over
, and
be the reflection of
over
, then
satisfies the problem conditions. (*)
It's easy to see with the given conditions that
Note that
so plugging into the third equation we have
Simplifying, this becomes
Of course, we can also "conjugate" this equation -- a nice way to do this is to note that if
then
whence
If
, then eliminating
, we get
The first case corresponds to (*) (since
uniquely determine
and
), the second corresponds to
(or equivalently, since
,
), and by symmetry, the third corresponds to
.
Otherwise, if , then we easily find
from the first of the two equations in
(we actually don't need this, but it tells us that the locus of working
is a line through the origin). It's easy to compute
and
, so
, and we're done.
Comment. It appears that taking the unit circle is nicer than, say
or
the unit circle (which may not even be reasonably tractable).
The problems on this page are copyrighted by the Mathematical Association of America's American Mathematics Competitions.
See Also
2011 USAMO (Problems • Resources) | ||
Preceded by Problem 2 |
Followed by Problem 4 | |
1 • 2 • 3 • 4 • 5 • 6 | ||
All USAMO Problems and Solutions |
The problems on this page are copyrighted by the Mathematical Association of America's American Mathematics Competitions.