0% found this document useful (0 votes)
3 views

8-Complex-Geometry-scan-1

The document discusses the use of complex numbers in solving geometry problems, emphasizing their advantages over traditional methods. It introduces key concepts such as addition, multiplication, and transformations involving complex numbers, while also presenting various problems for practice. The document aims to enhance understanding and application of complex geometry techniques for Olympiad enthusiasts.

Uploaded by

Paballo Mothapo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
3 views

8-Complex-Geometry-scan-1

The document discusses the use of complex numbers in solving geometry problems, emphasizing their advantages over traditional methods. It introduces key concepts such as addition, multiplication, and transformations involving complex numbers, while also presenting various problems for practice. The document aims to enhance understanding and application of complex geometry techniques for Olympiad enthusiasts.

Uploaded by

Paballo Mothapo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

~---------

) )
2 DIRKLAURJE

problems. To get full value from studying : .; material, you need to do


them all. If you get stuck, there are hints at the back to some of the
problems.
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST

1. COMPLEX NUMBERS
DIRK LAURIE
You know from Cartesian coordinate geometry that any point in the
plane can be represented by a pair of real numbers. Identify the point
(x, y) with the complex number z = x + yi where i is an imaginary
INTRODUCTION quantity with the curious property that i 2 = -1 (obviously i is not a real
number) which otherwise satisfies all the normal manipulation rules of
There are many ways to solve problems in geometry. Traditional syn- algebra. Therefore addition and multiplication of complex numbers are
thetic geometry often produces very short and elegant proofs if only one defined as follows: if z1 = x, + y, i and zz = xz + yzi, then
could find the correct construction to make, or the correct intermediate
hypothesis to prove. (1) z, + zz = (x1 +Y1l + (xz +yz)i;
Other methods, like trigonometry, vectors, transformation and inver- (2) z,zz = (x,xz -y,yz) + (x,yz +y,xz)i.
sion, occasionally provide short neat proofs. If all inspiration fails, one
can always resort to coordinate geometry and try to grind out a solution, Notation: in algebra, when you write two symbols together, as in ab, it
but this seldom actually works. You just lose too much when separating means multiplication of a and b, and that rule is followed here too, most
the coordinates. of the time. But since this booklet deals with geometry, we make one
In an actual Olympiad paper, there is not enough time to try out all exception: when all the symbols involve c7ital letters that are associ-
the various approaches. But there is good news for you: there is a single ated with points, then it may mean some gc.:..1netric object. For instance,
method that combines the virtues of trigonometry; vectors and coordi- A 1A2 , ABC, PQRS etc. usually denote lines, angles, polygons, etc. It will
nates. This all-in-one tool is analytical geometry using complex numbers. always be clear from the context when this happens.
If it doesn't work out with complex numbers, it probably would not have The formula for adding complex numbers is simply standard vector
worked out with any of those other alternative methods either. addition of points, when each point is identified with the vector from the
Complex geometry works because: the coordinates of a point are kept origin to that point. Geometrically, addition is translation: when a given
together, as in vector geometry; it is easy to describe translation, rotation point is added to each of a set of points, the set of points just gets moved
and reflection, as in transformation geometry; trigonometric identities around with no change in shape, orientation etc.
are in effect being applied all the time but you seldom need to be con- Multiplication looks complicated, but becomes easier to understand
scious of them; and all this is achieved by arithmetic operations on points. when you use polar coordinates , so that x = r cos 8 and y = r sin 8,
There are certain problems where it gives particularly neat and short so- where 8 is taken to be the anti-clockwise angle from the positive x-axis
lutions; you will meet several such here. Others, where it is clumsy and to the ray from the origin through (x, y). The modulus of a complex
cumbersome, also occur, but not here. number is defined as !zl = r. The polar angle when taken in the range
You need to acquire virtuosity in handling the complex geometry tools -180° < 8 :::; 180° is the argument of a complex number and denoted by
before they will become useful to you. There is no substitute for prac- arg z. Strictly speaking 8 should be in radians, but we simply define the
tice, and you should rework by complex numbers problems that you en- o symbol so that m = mn/180 and continue talking in terms of degrees.
0

counter elsewhere, even those that you could solve by other methods. The polar representation gives z = r( cos 8 + i sin 8) = re 6 i. Those
The problems in this booklet fall into three categories: exercises in who have learnt some calculus will know what e means, but that is not
the techniques, supplements to the theory, and actual competition level relevant here: we think of e 6 i as a convenient shorthand l}Ptation for
)
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST ' 3 4 DIRK LAURIE

~
Nothing is lost when you work with complex numbers and their con-
jugates instead of the x and y coordinates separately. You can always if
needed recover the real and imaginary parts of z by the formulas
(5) Rez=x= (z+z)/2,
(6) Im z = y = (z- z)/(2i.) = i.(z- z)/2,
FIGURE 1. Polar coordinates
and therefore it is never necessary to use x andy explicitly.
You should convince yourself of the identities
cos 8 + i.sin 8, easily remembered because the exponential law
z1 +zz =z1 +zz,
ee1 ie8zi = e(81 +9z Ji Z1Z2 = Z1Z2,
is satisfied (check it using trigonometry). Therefore so that the conjugate of an expression involving the four basic arithmetic
operations can be calculated by conjugating everything inside and then
(3) z1 zz = r1r2 (cos (81 + 82 ) + i.sin (8 1 + 8 2)) = r 1r 2e( 91 +Bz li; doing the arithmetic. An expression such as zz which remains unchanged
(4) zdzz =rdrz(cos(81-8z)+i.sin(81 -8 2)) =rdr 2e(e 1- 92 li. when conjugated, must always be real.
There are several important reasons for introducing conjugates:
In words: to multiply two complex numbers, you multiply the radii and
add the angles. If one of the points (say z 1) is on the unit circle r = 1, • It simplifies division so you need only divide by real numbers,
since
then the effect of multiplication by z 1 is rotation of the other point by the Z1 Z1Z2 Z1Z2
angle 8 1 around the origin. If one of the points (say z2 =c) is real, then zz - zzzz - --;z·
the effect of multiplication is scaling by a factor c.
A word of caution about the argument: equations like • Take a polynomial equation with real coefficients, e.g. z4 +z+ 1 =
0. Conjugate everything to obtain z4 +z+ 1 = 0. You get the same
arg(z 1z2 ) = argz 1 + argzz equation as before with z as the unknown. Therefore every root
arg(zdzz) = argz 1 - argzz either satisfies z = z (i.e. z is real) •.r there is another root z.
• It is inconvenient to work with equations in which some quanti-
are only true when the right-hand side lies in the correct range. Other- ties are complex but others are restricted to be real, such as:
wise all we can say is that the equation is true modulo 360°.
When two angles differ by a multiple of 360°, they look the same when z = z 1 + t(z2 - z 1) where tis real.
sketched and are represented by the same complex number, but they You easily lose track of the extra information, and anyhow the
may behave differently in computation. For example, one can solve the variable t is an unnecessary extra symbol. Instead, write the
equation z5 = 1 as follows: let Z = COS 8 + i. sin 8, then z5 = COS 58 + statement "(z- z1)/(z2 - z 1) is real" in the form
i. sin 58 = 1 = cos 0 + i. sin 0. Therefore 8 = 0 gives a solution z = 1. But it Z-Z1 z-z1
is also true that 1 = cos 360° + i. sin 360°) which leads to e = 72°) etc. Zz -Z1 z2 -z 1
1.1. Conjugates. The conjugate z of a complex number z is defined by in which all the quantities may be as complex as they please.

X+ iy =X- l1J.
Problem 1. The "n-th root of unity" is defined by
Wn = e2ni/n.
Note that r 2 = lzl 2 = zz. The geometrical interpretation is that z is the
reflection of z in the x-axis. Show that:

----· ~------- ---- ------ ------------ - - - - - - - - - - - ___l _ _ _ _ _ _ _


) l
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 6 DIRK LAURIE

(1) the numbers 1, Wn, w~, ... , w~- 1 lie at the vertices of a regular
Translation 0' 0 Reflection ~<\

~ w~e'' ~
polygon;
(2) 1 + Wn + w~ + .. · + w~- 1 = 0;

ZHz+O'-~ ~ zHQ+w (z-Q)~


2
(3) when n = 2m where m is an odd number, then
1 -Wzm +w~m -w~m + ·· · +w.f,:;_- = 0.
1

Problem 2. For all complex numbers z = x + iy not on the negative real


axis, we define /Z = u + vi so that u > 0; the other solution to w 2 = z Rotation A<\ Scaling w = 1.2

v~~ ~0
is obviously w = -/Z. Show that u = J(r + x)/2 and v = ±J(r- x)/2,
where v has the same sign as y.
= eie

~
w
Problem 3. If tan 48 = l, calculate all possible values of tan 8.
1

zHO+w(z-0)
2. TRANSFORMATIONS
FIGURE 2. Transformations and complex arithmetic. Ro-
We have already seen that translation, rotation, scaling and reflection tation and scaling have the same formula, but rotation
are readily expressible in terms of complex arithmetic, so you can do requires lwl = 1 where scaling requires w to be real.
transformation geometry by complex numbers. This is a powerful tool in
finding quick solutions to some problems.
The basic principle is that the linear transformation z H az + b is a changes it to wAwB = AWwB = AB because Ww = lwl 2 = 1; and A/B
combination of translation, rotation and scaling, which are all operations is invariant under rotation and scaling, since the transformation z H cz,
that preserve lengths and angles. Many problems are much easier when where c is any complex number, takes it to itself.
important points lie on the origin or on the coordinate axes. If we do Invariant expressions are important in deriving formulas. For exam-
not have the freedom to choose the coordinate system, we can always ple, the area of a rectangle with vertices at D(O, 0), A( a, 0), B(O, b) and
transform to the one we like, do the problem in an easy way, and then C( a, b) is ab. There are several complex expressions that all give this
transform the result back to the original system. result, among them -iAB and iBA. But -iAB is not invariant under ro-
Here is a simple but important example. tation, whereas iBA is. Therefore iBA gives the area of any rectangle
DACB when D = 0. Another possible expression is i(B- D)(A- D),
Perpendicular from point to line. It is easy to find the foot D of the which is invariant under translation and rotation, and therefore gives the
perpendicular from C to the line AB when A = 0 and B is real: the area of any rectangle DACB. Note that the area of DBCA comes out
answer is D = Re C. The linear transformation that takes A to 0 and B to negative: in the next section we discuss wh<J.r that means.
1 is z H (z- A)/(B- A). Therefore in the general case,
Problem 4. Write expressions for the following operations:
D-A C-A
B -A= ReB -A (1) Rotate z clockwise by 20° around the point P.
(2) Reflect the point A in the line through P and Q.
C-A
(7) ~ D =A+ (B -A)Re B -A
Problem 5. Show that another formula for dropping a perpendicular to a
Some complex expressions are invariant under certain transforma- line is
C-A
tions. For example, A - B is invariant under translation, because the D = C- i(B -A) Im B _A.
transformation z H z+w changes it to (A+w)- (B +w) =A- B; AB is
invariant under rotation, since the transformation z H wz with lwl = 1 Problem 6. It is known from transformation geometry that

l__ _l__~
)
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST ' 7 8 DIRK LAURIE

~o,cBA<~A
(1) Reflection in the line AB followed by reflection in the line CD is
equivalent to a rotation.
(2) Rotation bye 1 around Z1 followed by~fy ez around Zz is
equivalent to a single rotation. ~c
Prove both statements by calculating the centre and angle of rotation in
[6ABC] > 0, [6CBA] < 0
each case.

3. GEOMETRICAL MEASUREMENTS FIGURE 3. Positive angles and areas always work with
points taken anti-clockwise.
Lengths are always positive, but sides are have direction: AB stands
for the directed line segment from A to B, which is the vector B- A. The
length of AB is IB- AI. the perimeter of a triangle (i.e .. walk around it along the directed sides
For any point P, it is a matter of simple arithmetic that until you get back to where you started), the enclosed area is positive if
you have gone through an angle of 360° and negative if you have gone
A - P = m {::::=} p = mA + nB
through -360°. If you are too lazy to walk, just stand on the perimeter
P-B n m+n
looking forwards: the area of the triangle is positive if its interior is to
When Plies on the line segment AB, it therefore divides AB in the ratio your left.
m:n. These two ways of telling the sign of an area apply to any non-inter-
Angles and areas may be negative. A positive angle is measured in secting figure, even curved ones. They also apply to a figure with holes
an anti-clockwise direction, a negative angle in a clockwise one. The in, as long as the perimeter of each hole is specified so that the interior of
directed angle ABC is the amount by which BC needs to be rotated until the figure lies the same way (always or never to the left) on each directed
it lies on BA. It is usually convenient to think of a reflex angle e as a line segment.
non-reflex angle e ± 360° of opposite sign. The area of a triangle is calculated by the formula
An angle is calculated by
(8) [ABC]= 1 Im(AC + BA + CB).
ABC= arg(A- B)/(C- B),
You will be asked to prove this formula in the exercises. Alternative forms
not by arg(A- B)- arg(C- B), in order to make sure that we get an that are occasionally more convenient are:
angle in the correct range. Generally
CBA =-ABC (9) [ABC]=~ (C(B- A)+ B(A- C)+ A(C- B))
except when both angles are 180° : the definition of 'argument' does not (10) = 1Im ((B- A)(C- AI).
1il> allow - 180° .
The sign of the area of a figure depends on how its perimeter is spec- If CAB = 90°, (B- A) ( C- A) is purely imaginary and we have
ified. We start with triangles. If the angle ABC is such that ABC > 0, i- -
[ABC]= --(B- A)(C- A).
then also its area [ABC] = [BCA] = [CAB] > 0; but [ACB] = [CBA] = 2
[BAC] = -[ABC] < 0. (The sine area formula [ABC] = 1bcsinABC The area formula generalizes to an arbitrary non-intersecting polygon
thus remains true even when ABC< 0. But you won't need this formula: A1A2 ... An:
there are easier ways to calculate areas.)
1
2L
n -
There are two other ways to tell the sign of an area. The notation (11) [A1A2 ... An]= Im(AkAk-1),
ABC indicates that the directed sides are AB, BC and CA. If you traverse k=1

l_
~ ~
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 9 10 DIRKIAURJE

where Ao = An. If the polygon is convex and the origin is inside, you In complex geometry, it is convenient to define triangles ABC and
can immediately recognize the formula as the sum of the areas of the PQR as similar (written .6.ABC - .6.PQR) only when they can be made
triangles Ak-1 Ak 0. If not, our convention about negative areas ensures homothetic by rotating one of them. This requires that they are traversed
that the formula is still correct. in the same direction, and one can be made to coincide with the other
A useful application of signed areas is area coordinates. Let P be any by some combination of rotation, translation and scaling. There must be
point, and define certain complex numbers w =1= 0 and z such that
a= [PBC]/[ABC], (3 = [PCA]/[BCA], ;y = [PAB]/[CAB]. A= z+wP, B = z+wQ, C = z+wR.
Then it can be shown that Eliminating w and z from these three equations, we find that
(12) P = aA + (3B + yC, where a, (3, y E IR and a+ (3 + y = 1. A-B P-Q
.6.ABC- .6.PQR <===? C _ B = R _ Q.
Problem 7. Prove the formula:for the area of a triangle.
Problem 8. Verify (12) using reversible steps in your argument, thereby If they are traversed in opposite ways, we need to reflect one of them
showing that ifP is given by (12), then a, (3 andy are uniquely determined. to make them the same. So the test becomes
A-B P-Q
4. GEOMETRIC RELATIONS INVOLVING ANGLES .6.ABC -rev .6.PQR <===? - C B = =--==·
- R-Q
To test in general whether two directed angles are equal, the require- In Euclidean geometry we usually define similarity of triangles by
ment is that
.6.ABC Ill .6.PQR {===} .6.ABC- .6.PQR or .6.ABC -rev .6.PQR.
~ ~ A-B;D-E
ABC= DEF <===? C _ B F _ E is real and positive.
This concept is seldom required in complex geometry.
The 'and positive' part is annoying since it becomes difficult to manip- Triangles are congruent if one can be made to coincide with the other
ulate the relation. It may be dropped when we do not need to distinguish by some combination of rotation, translation and reflection: like being
between angles e and e + 180°. similar, but with scale factor 1. This gives
Two lines AB and CD are parallel when arg( A - B) = arg( C - D) or C-A B-A
arg(A- B) = arg(C- D)± 180°; combine the cases by observing that .6.ABC::::: .6.PQR <===? R _ p = Q _ p =w with lwl = 1.
they both say (A - B) I (C - D) is real. Therefore
As in the case of similar triangles, for full compatibility with the tradi-
(13) AB II CD <===? (A- B)I(C- D)= (A- B)I(C- D). tional definition of congruence one needs to define a '=rev relation. How-
Three points A, B and Care collinear when AB II BC. ever, congruent triangles are not important in complex geometry, since
Two lines AB and CD are perpendicular when arg(A- B) = arg( C- the things we use them for (proving equality of lengths and angles) can
D) ± 90°; combine the cases by observing that they both say (A-B) I (C- be done directly once we have the points.
D) is purely imaginary. Therefore Problem 9. Three equilateral triangles ABP, CDP and EFP have a common
(14) AB _!_CD <===? (A- B)I(C- D)= -(A- B)I(C- D). vertex P such that BPC+DPE+FPA = 180°. l, M and N are the midpoints
Two geometric figures are homothetic with centre A if one can be made of BC, DE and FA respectively. Prove that lM N is an equilateral triangle.
[POTW]
to coincide with the other by scaling with fixed point A. In other words,
there exists a real constant c such that for any point z on the first figure, Problem 10. Prove Napoleon's theorem, which says that if you draw equi-
there is a point z' on the second figure so that lateral triangles on all three sides of an arbitrary triangle, their centroids
z' -A= c(z-A). form an equilateral triangle.

!
--~~-~- ------ ----- __L -------
il
COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 11 12 DIRKLAURJE

Problem 11. Three circles, with radii p, q, r, and centres A, B, C respec- a line. Conjugate everything to obtain az + bz = c. This equation only
tively, touch one another externally at points D, E, F. Prove that the ratio describes a line if it says exactly the same as the original one. Therefore
of the areas of 6DEF and 6ABC equals we need b/a = a./b, i.e. lal = lbl, and uc = be. If the first of these
2pqr conditions is not satisfied, we can solve the two equations for z; if the
(p+q)(q+r)(r+p)" first is satisfied but not the second, the two equations are in conflict.

[SAMO 1995] 5.2. Circles. The equation of a circle with radius r and centre Q is very
easy
5. EQUATIONS OF LINES AND CIRCLES lz-QI 2 =r 2 .
In complex geometry, all equations are written in terms of z and z, We will denote this circle by K(Q, r). Multiplying out, we obtain
not in terms of x and y. This may not seem to be a great advantage, (17) zz- Qz- Z:Q = r 2 - QQ.
but there is an inherent symmetry between a complex number and its
conjugate that does not exist between the coordinates themselves. Most From this you can see that an equation of the form zz + az + bz + c = 0
of the equations are self-conjugate: when you conjugate everything, you describes a circle if and only if b = a, c is real, and lbl 2 - c > 0.
just get the original equation again. When you have two equations, you To find the circumcircle of 6ABC, we substitute the three vertices into
solve only for z. (17) and subtract one equation from both others to obtain

5.1. Lines. A general point z on the line AB is collinear with A and B, Q(A- B)+ Q(A- B)= AA- BB;
so from (13) we get Q(A-C) +Q(A- C) =AA- CC.
(z- A)/(A- B)= (z- A)/(A- B). Eliminating Q and regroup:
Putting s =A- B, we obtain
Q = A(B- C)A+ B(C -A)B + C(A- B)C.
(IS)
z z A A (B- C)A + (C- A)B +(A- B)C
s s
s s
The radius is found from e.g. r 2 =(A- Q)(A- Q), which simplifies to
In general, we denote a line through A parallel to s by e(A II s).
It is sometimes more convenient to descibe a line in terms of a vector (A-B)(B-C)(C-A) I
perpendicular to it. Put w =is in (IS) to obtain r = I (B- C)A+ (C -A)B+ (A- B)C
z
-+-=-+-.
z: A A To find the incircle, the easiest method is by area coordinates. When P
w w w w is the incentre I, the triangles PBC, PCA and PAB have equal height and
In general, we denote a line through A perpendicular tow by e(A _1_ w). therefore the area is proportional to the bases. We obtain
When f(A _1_ w) does not pass through zero, it contains a point p AlB - Cl + BIC- AI .J_ CIA- Bl
closest to zero, so that p is perpendicular to f(A j_ w). Choosing A = I= .. .
IB - Cl + IC- AI + lA- Bl
w = p, we get
z z The excircles are found similarly. E.g. for the excircle opposite to A, the
(16) -+-
p 'P =2 . triangle PBC is traversed in the opposite direction and therefore has neg-
ative area proportional to -IB- Cl. We obtain
This equation fully characterizes a line in terms of only one point.
When we arrive at an equation of the form az + bz = c by other -AlB - Cl + BIC- AI +CIA- Bl
means, it is useful to know whether the equation does in fact describe IA = -IB- Cl + IC- AI+ lA- Bl etc.

---- _____! -·-----


") ~

COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 13 14 DIRK lAURIE

Problem 12. Find the radius of the incircle by calculating II- Rl. Manipu- The answer can be simplified quite a bit by noting that there are many
late the result to obtain an expression that is symmetric in A, B and C. subexpressions of the form z - z = 2i Im z. We obtain
5.3. Tangents. Suppose that A is on the circle K(Q, r). If Q = 0, finding X= (B- D) Im(CA)- (A- C) Im(DB)
the tangent is trivial, since A is the least point on the tangent. This would Im ((B- D)(A- C))
give z/A + z/A = 2. To move the origin to 0, do a translation:
= A Im(BD) + B Im(CA) + C Im(DB) + D Im(AC)

I
z-Q z-Q Im(AD + BA + CB +DC)
(18)
A-Q +A-Q =2.

Suppose that A is on the circle K(Q, r), and we wish to find the centre
(19) = t; Ak Im(Ak+ 1Ak-1) t; Im(AkAk-1) ,
4 4

of the circle K(Q', r') of radius r' that touches K(Q, r) at A. The easiest
method is by homothety: if z' is a point on K(z', r'), then there is a point where Ao = A4 and A 1 = As. The last form of the formula is almost
z on K(Q, r) such that (z'- A)/r' = ±(z- A)/r (the plus sign applies simple enough to remember after all: the denominator is just the area of
when the circles touch internally and the minus sign when they touch the quadrilateral.
externally). The homothety applies to the whole figure, including the In an actual calculation, one would hope that the four given points are
centre. Therefore Q' =A± (r' /r)(Q- A). not totally independent, and that further simplifications occur.

6.2. Intersection of line and circle. This one is tricky. You can elimi-
6. CONSTRUCTIONS nate z from the two equations to get a quadratic in z, but this quadratic al-
ways has a solution (since we are working with complex numbers) while
Constructions involve setting up and solving equations. There is not
the line may be missing the circle. To be safe, first calculate the point P
much point in trying to remember formulas even in simple, common
cases such as the intersection of two lines. You must be able to apply on the line that is closest to the centre of the circle K(Q, r). If IP - Q I > r,
the basic principles to the case at hand. there is no intersection.

6.3. Intersection of two circles. First find the radical axis, which is the
6.1. Intersection of diagonals of quadrilateral. The problem is of course line through the intersection points of K(Q1, r 1) and K(Qz, rz). Subtract
the same as finding the intersection of two lines, but it is easier to think the equation for the second circle from that of the first. The quadratic
in terms of a quadrilateral ABCD. Let X be the intersection of AC and terms cancel, and we obtain
BD. In order to retain the symmetry of the four given points, set up the
collinearity equations in the form (20) (Qz- Q1 )z+ (Q 2 - Q 1)z = rf- r~ -IQ1I 2 + 1Qzl 2

X-A X-A which is the equation of a line. Now find the intersection of either circle
x-c=x-c with the radical axis.
The radical axis is the common tangent '"hen the circles touch. It is
{==} X(A- C) -X( A- C) = AC- AC, still defined by (20) even when the circles are disjunct, but its geometri-
and similarly cal meaning is then more complicated.
X(C- D) -X(C- D)= CD- CD.
7. ADVANCED GEOMETRIC RELATIONS
Solving for X, we find
7.1. Concurrent lines. To test in general whether lines are concurrent,
X= (AB- AB)(C- D)- (CD- CD)(A- B) you calculate the intersection of two of the lines (see Section 6) and then
(A- B)(C- D)- (C- D)(A- B) test whether that point is collinear with the two remaining points, which

I
_ ___j
~-------~
- ~ ----~ - - ·----- -------- - - - - - - - - - ---~-
~ ~

COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 15 16 DIRK lAURIE

can be quite tedious. There is however a neat shortcut when the lines are Supplementary angles arise when
in closest-point form.
arg(A- B)/(C- B)+ arg(C- D)/(A- D)= 180°
Put a = 1/p in (16) to obtain
(A- B)(C- D)) o
(21) az+ iii= 2. {===} arg ( (C- B)(A- D) = 180 .
When we speak of "the line a," we mean the line given by equation (21). The two cases can be combined by saying that
Equation (21) remains unchanged if we exchange a (which stands for
A-B ;c-B
a line) and z (which stands for a point). Therefore three equations saying (25) A, B, C and 0 are cocylic A_
{===} C_ is real.
0 0
that three lines a, b and c all pass through a point z look exactly like three
equations saying that three points a, b, and c all lie on a line z. So the The expression in (25) is called the cross-ratio of the four points and
three lines a, b and c are concurrent if and only if the three points a, b denoted by (A, C; B, D).
and c are collinear. But we already have a test for collinearity by using Note that the degenerate case where four points are collinear also
equation (13). This gives counts as cocyclic. (Think, if you will, that the circle has infinite radius.)
a-b c-b Problem 14. ABC is a triangle with sides 1, 2, and J3. Determine the
(22) Lines a, b and c are concurrent {===} - - - = --=. smallest possible area of an equilateral triangle with one vertex on each of
a:-b c-b
the sides of ABC.
When the lines AD, BE and CF join each vertex of 6ABC to a point [SAMO 1996]
on the opposite side, you can use the complex form of Ceva's theorem,
which states: Problem 15. The convex quadrilateral ABCD has AC j_ BD and the
When D, E and Fare on BC, CA and AB respectively; then perpendicular bisectors of AB and CD meet at a point P inside ABCD.
Show that ABCD is cyclic if and only if [ABP] = [COP].
0-BE-CF-A
(23) AD, BE and CF are concurrent {===} _ C E _A F _ B = -1. [IMO 1998]
0
Menelaus' theorem looks very similar: When 0, E and Fare on BC, CA Problem 16. Circles k 1 and k 2 are drawn so that k 2 passes through the
and AB respectively; then centre of k 1 • The circles intersect in points A and B. A third circle k 3 touches
k 1 inC and k 2 in 0 so that k 1 and k 2 are inside k3. AB is extended to meet
0-BE-CF-A k3 in E and F. The lines DE and OF intersect kz respectively in G and H.
(24) 0, E and F are collinear {=:=} _ C E _ A F _ B = 1·
0 Prove that GH is tangent to k 1 .
Problem 13. Given {J = D-C'D-B 1l = E-C and rh = F-A such that fJT]rh =
[IMO 1999]
E-A 'I' F- B 'I'
-1, calculate ex, j3 andy such that the intersection P of AD, BE and CF is Problem 17. A, B, CD, E and F lie in that order on the circumference of
given by a circle. The chords AD, BE and CF are concurrent. P, Q and R are the
P = ocA + j3 B + yC. midpoints of AD, BE and CF respectively. 1Wo further chords AG II BE and
AH II CF are drawn. Prove that 6PQR Ill 6DGH.
7.2. Cocylic points. Four points A, B, C and 0 are cocylic if and only if [SAMO 1998]
either ABC= ADC, or ABC and CDA are supplementary. Equal angles
arise when
arg(A- B)/(C- B)- arg(A- D)/(C- D)= 0

{===}
(A- B)(C-
arg ( (C- B)(A- D)
D)) = O.
) )

COMPLEX GEOMETRY FOR THE OLYMPIAD ENTHUSIAST 17

HINTS
INDEX
Problem 5: Which linear transformation maps C to 0 and B- A to addition, 2 invariant, 6
1? angle
Problem 7: Show that the formula is invariant under translation directed, 5, 7 linear transformation, 5
and rotation, and that it therefore suffices to show its validity reflex, 7 Menelaus' theorem, 15
angles
when A = 0 and B is real. modulus, 2
equal, 8 multiplication, 2
Problem 9: Take the origin at P and show that MN is LN rotated area, 7
around N through 60°. polygon, 8 Napoleon's theorem, 10
Problem 10: The centroid of L-ABC is !(A+ B +C). rectangle, 6
sign of, 7 parallel, 9
Problem 14: Take A = 0, B = 1 and C = 2w, where w = e 60 oi. perpendicular, 5, 9, 11
triangle, 7
Then D + D = 2, E/w = E/w and F =F. Find the side length argument, 2 polygon, 4, 8
6 D EF in terms of F. polynomial, 4
Problem 15: Calculate the complex number P, and alsoP', which calculus, 3
radians, 2
satisfies [ABP'] =[COP']. Try to factorize P- P'. capital letters, 2
Ceva's theorem, 15 radical axis, 14
Problem 16: Choose k1 = K(1,r), k2 = K(O, 1), k3 = K(Q,s). Cal- circumcircle, 12
real, 4
culate Q in terms of r and s. You need to prove that ReG reflection, 3
cocylic, 15
ReH = 1 -T. collinear, 9, 11
root of unity, 4
concurrent, 14 rotati<Jn, 3, 5
congruent, 10 SAMO
conjugate, 3 1995, 10
coordinates 1996, 16
area, 8, 12 1998, 16
Cartesian, 2 scaling, 3, 5
polar, 2 self-conjugate, 10
cross-ratio, 15 side
directed, 7
degrees, 2 similar, 9
direction, 6 square root, 5
division, 4
tangent, 12
equation two circles, 12
circle, 11 transformation geometry, 5
line, 11 translation, 2, 5
excircles, 12 traverse, 7
exponential law, 3
unit circle, 3
homothetic, 9
vector, 2
imaginary, 2, 4
IMO
1998, 16
1999, 16
incenrre, 12
incircle, 12
18

-- ------ --------

You might also like