Isogonal conjugation with respect to a triangle Darij Grinberg version 23 September 2006 1.

Trivialities on isogonal lines The aim of this note is to provide a rather detailled and general introduction into isogonal conjugation with respect to a triangle. No new results of the author will be presented, but some results published elsewhere will be proven in a (presumably) new way. Throughout this work, we will make use of directed angles modulo 180 : This is the kind of angles referred to as "directed angles" in [1], 1.7, and we refer the reader to [1] for their basic properties. (Also, [2], [3] and [4] provide introductions to this type of angles.) Two preliminary conventions are to be made at rst: "Wrt" is an abbreviation for "with respect to". The A-altitude of a triangle ABC will mean the altitude of triangle ABC issuing from its vertex A: Similarly, the A-median of triangle ABC will mean the median of triangle ABC issuing from its vertex A:

U w

p w' u q

Fig. 1 We start with a simple property of lines (Fig. 1): Theorem 1. Let u and v be two lines which intersect at an Euclidean (i. e. not innite) point P: Let p and q be two more lines through the point P: Let w and w0 be the two angle bisectors of the angles formed by the lines u and v: Then, the following four assertions A1 ; A2 ; A3 and A4 are pairwisely equivalent: Assertion A1 : We have ] (u; q) = ] (v; p) : Assertion A2 : We have ] (v; q) = ] (u; p) : Assertion A3 : The lines p and q are symmetric to each other wrt the line w: Assertion A4 : The lines p and q are symmetric to each other wrt the line w0 : Proof of Theorem 1. This proof is almost trivial; we are giving it here only for the sake of completeness. First, if assertion A1 holds, i. e. if ] (u; q) = ] (v; p) ; then ] (v; q) = ] (v; u)+] (u; q) = ] (v; u)+( ] (v; p)) = (] (v; p) ] (v; u)) = ] (u; p) ; so assertion A2 also holds. Similarly, we show the converse: if assertion A2 holds, then assertion A1 also holds. Thus, the assertions A1 and A2 are equivalent. Now we will prove the equivalence of assertions A1 and A3 : Since the line w is an angle bisector of one of the angles between the lines u and v; we have ] (u; w) = ] (v; w) : Now, assertion A1 states that ] (u; q) = ] (v; p) ; this rewrites as ] (u; w)+] (w; q) = (] (v; w) + ] (w; p)) ; what, in view of ] (u; w) = ] (v; w) ; simplies to ] (w; q) = ] (w; p) : But since the lines w; p; q all pass through the point P; this equation is equivalent to stating that the lines p and q are symmetric to each other wrt the line w: This, however, is assertion A3 : Thus we have shown that the assertions A1 and A3 are equivalent; similarly we can prove the assertions A1 and A4 to be equivalent. Hence, the equivalence of all four assertions A1 ; A2 ; A3 and A4 is established, and Theorem 1 is proven. Based on Theorem 1 we make a denition: If one of the four assertions A1 ; A2 ; A3 and A4 holds - and therefore, according to Theorem 1, the three others hold as well -, then we say that the line q is isogonal to the line p wrt the lines u and v: For two lines p and q through the point P; we can easily see that the line q is isogonal to the line p wrt the lines u and v if and only if the line p is isogonal to the line q wrt the lines u and v: (In fact, the line q is isogonal to the line p wrt the lines u and v if and only if ] (u; q) = ] (v; p) ; this is equivalent to ] (v; p) = ] (u; q) ; and this holds if and only if the line p is isogonal to the line q wrt the lines u and v:) Hence, the assertions "the line q is isogonal to the line p wrt the lines u and v " and "the line p is isogonal to the line q wrt the lines u and v " are equivalent; thus, instead of any of these assertions, we can simply say that "the lines p and q are isogonal to each other wrt the lines u and v ". Instead of saying "isogonal wrt the lines u and v ", we will often say "isogonal wrt the angle U P V ", where U is a point on the line u (distinct from P ) and V is a point on the line v (distinct from P ). For each line p through the point P; there exists one and only one line q through the point P which is isogonal to the line p wrt the lines u and v; in fact, this line q must satisfy the equation ] (u; q) = ] (v; p) ; and this holds for one and only one line q through the point P (this line can be constructed as a line through a given point which forms a given angle with another given line). This line q which is isogonal to 2

the line p wrt the lines u and v is called the isogonal (or isogonal line) of the line p wrt the lines u and v; or, equivalently, the isogonal (or isogonal line) of the line p wrt the angle U P V: 2. Isogonals and perpendicular bisectors The next properties of isogonals we are going to show are not much harder to prove, but turn out to be of remarkable usefulness: Theorem 2. Let u and v be two lines intersecting at an Euclidean point P: Let T be a point in the plane. a) Let X and Y be the orthogonal projections of the point T on the lines u and v: Then, the line XY is perpendicular to the isogonal of the line P T wrt the lines u and v: (See Fig. 2.)

r v P Y

Fig. 2 b) Let X 0 and Y 0 be the re ections of the point T in the lines u and v: Then, the perpendicular bisector of the segment X 0 Y 0 is the isogonal of the line P T wrt the lines u and v: (See Fig. 3.)

X'

r v P

Y'
Fig. 3 Proof of Theorem 2. Let r be the isogonal of the line P T wrt the lines u and v: Then, ] (u; r) = ] (v; P T ) and ] (v; r) = ] (u; P T ) : a) (See Fig. 4.) Since ]P XT = 90 and ]P Y T = 90 ; the points X and Y lie on the circle with diameter P T: Thus, ]Y XP = ]Y T P; so that ] (XY ; u) = ] (T Y ; P T ) : But T Y ? v yields ] (T Y ; v) = 90 ; so that ] (XY ; u) = ] (T Y ; P T ) = ] (T Y ; v) + ] (v; P T ) = 90 + ] (v; P T ) : Hence, ] (XY ; r) = ] (XY ; u) + ] (u; r) = (90 + ] (v; P T )) + ( ] (v; P T )) = 90 : Thus, the line XY is perpendicular to the line r; that is, to the isogonal of the line P T wrt the lines u and v: This proves Theorem 2 a).

X T u

r v P Y

Fig. 4 b) (See Fig. 5.) Since X 0 is the re ection of the point T in the line u; we have P X 0 = P T ; similarly, P Y 0 = P T: Thus, P X 0 = P T = P Y 0 ; what entails that the point P is the center of the circle through the points X 0 ; T; Y 0 : Thus, on the one hand, the central angle theorem for directed angles modulo 180 yields ]P Y 0 X 0 = 90 ]X 0 T Y 0 : On the other hand, P X 0 = P Y 0 implies that P lies on the perpendicular bisector of the segment X 0 Y 0 : Since Y 0 is the re ection of the point T in the line v; we get ] (P Y 0 ; v) = ] (P T ; v) and T Y 0 ? v; the latter yields ] (v; T Y 0 ) = 90 : Similarly, ] (u; T X 0 ) = 90 : Therefore, ] (P Y 0 ; X 0 Y 0 ) = ]P Y 0 X 0 = 90 ]X 0 T Y 0 = 90 ] (T X 0 ; T Y 0 ) = ] (v; T Y 0 ) = ] (v; T X 0 ) = ] (v; u) + ] (u; T X 0 ) = ] (v; u) + 90 : Consequently, ] (v; X 0 Y 0 ) = ] (P Y 0 ; X 0 Y 0 ) ] (P Y 0 ; v) = (] (v; u) + 90 ) ( ] (P T ; v)) = (] (P T ; v) + ] (v; u)) + 90 = ] (P T ; u) + 90 ; so that ] (r; X 0 Y 0 ) = ] (v; X 0 Y 0 ) ] (v; r) = (] (P T ; u) + 90 ) = (] (P T ; u) + 90 ) ] (P T ; u) = 90 : 5 ( ] (u; P T )) ] (T X 0 ; T Y 0 )

This means that the line r is perpendicular to the line X 0 Y 0 : Now, the perpendicular bisector of the segment X 0 Y 0 is also perpendicular to the line X 0 Y 0 : Hence, the perpendicular bisector of the segment X 0 Y 0 is parallel to the line r: But since the perpendicular bisector of the segment X 0 Y 0 and the line r have a common point (namely P ), they can only be parallel if they coincide. Hence we see that the perpendicular bisector of the segment X 0 Y 0 coincides with the line r; that is, with the isogonal of the line P T wrt the lines u and v: This proves Theorem 2 b) and thus completes the proof of Theorem 2.

X'

T u

r v P

Y'
Fig. 5 Remark. (See Fig. 6.) As X is the orthogonal projection of the point T on the line u; whereas X 0 is the re ection of T in u; we see that X is the midpoint of the 0 segment T X : Similarly, Y is the midpoint of the segment T Y 0 : Thus, the line XY is a midparallel in triangle X 0 T Y 0 and thus parallel to its side X 0 Y 0 : Hence, the assertion that X 0 Y 0 ? r (the crucial assertion in the proof of Theorem 2 b)) is equivalent to the assertion that XY ? r (this is the assertion of Theorem 2 a)). This shows that Theorems 2 a) and 2 b) can be derived from each other.

X'

X T u

v P

Y'
Fig. 6 Note that Theorem 2 will not only turn out useful to us in our study of isogonal conjugates, but it can also be applied to olympiad problems like the IMO 2004 problem 5 ([8], post #2). 3. Isogonal conjugation wrt triangles

C A

Fig. 7 Now we show the rst serious result on isogonals, the isogonal conjugate theorem: Theorem 3. Let ABC be a triangle and P a point (distinct from the vertices A; B; C). Then, the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA concur at one point Q: (See Fig. 7.) It has to be noticed that this theorem is formulated for the projective plane - this means that the point P can be an Euclidean or an innite point, and that the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA can concur at an Euclidean or at an innite point as well. Proof of Theorem 3.1 We distinguish between two cases: Case 1: The point P is an Euclidean (i. e. not innite) point. (See Fig. 8.) Let X 0 ; Y 0 ; Z 0 be the re ections of the point P in the lines BC; CA; AB:
The following proof is basically the proof given in [9], apart from the dierence that [9] doesn t use directed angles.
1

Since Y 0 and Z 0 are the re ections of the point P in the lines CA and AB; according to Theorem 2 b) it follows that the perpendicular bisector of the segment Y 0 Z 0 is the isogonal of the line AP wrt the lines CA and AB; that is, the isogonal of the line AP wrt the angle CAB: Similarly, the perpendicular bisectors of the segments Z 0 X 0 and X 0 Y 0 are the isogonals of the lines BP and CP wrt the angles ABC and BCA: Thus, the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA are the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 and thus concur at one point - at the circumcenter of triangle X 0 Y 0 Z 0 : This proves Theorem 3 in Case 1.

B X' Z'

P Q

C A

Y'
Fig. 8 Case 2: The point P is an innite point. Then, the lines AP; BP; CP are parallel. (See Fig. 9.) Let the isogonal of the line AP wrt the angle CAB meet the circumcircle of triangle ABC at a point Q1 (apart from A). Since the lines AP and AQ1 are isogonal to each other wrt the angle CAB; we have ] (CA; AQ1 ) = ] (AB; AP ) : Since Q1 lies on the circumcircle of triangle ABC; we have ]AQ1 C = ]ABC; thus

] (AQ1 ; CQ1 ) = ] (AB; BC) : Hence, ] (CA; CQ1 ) = ] (CA; AQ1 ) + ] (AQ1 ; CQ1 ) = ] (AB; AP ) + ] (AB; BC) = (] (AB; AP ) ] (AB; BC)) = ] (BC; AP ) : But AP k CP yields ] (BC; AP ) = ] (BC; CP ) ; and thus we obtain ] (CA; CQ1 ) = ] (BC; CP ) : This means that the line CQ1 is the isogonal of the line CP wrt the angle BCA: In other words, the point Q1 lies on the isogonal of the line CP wrt the angle BCA: Similarly, the point Q1 lies on the isogonal of the line BP wrt the angle ABC: We already know that the point Q1 lies on the isogonal of the line AP wrt the angle CAB: Thus, the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA concur at one point - namely, at the point Q1 : Herewith we have not only proved Theorem 3 in Case 2, but we have also shown that in this case - i. e. in the case when the point P is innite -, the point of intersection of the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA lies on the circumcircle of triangle ABC (in fact, this point of intersection is our point Q1 and lies, by its denition, on the circumcircle of triangle ABC).

Q1 B

C A

10

Fig. 9 Now Theorem 3 is completely proven. Based on Theorem 3, we introduce a notion: The point of intersection of the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA is called the isogonal conjugate of the point P wrt the triangle ABC: If Q is the isogonal conjugate of a point P wrt the triangle ABC; then the line AQ is the isogonal of the line AP wrt the angle CAB: This means that the lines AP and AQ are isogonal to each other wrt the angle CAB: This, in turn, shows that the line AP is the isogonal of the line AQ wrt the angle CAB: Similarly, the lines BP and CP are the isogonals of the lines BQ and CQ wrt the angles ABC and BCA: Now, the point P is the point of intersection of the lines AP; BP; CP; so it therefore must be the point of intersection of the isogonals of the lines AQ; BQ; CQ wrt the angles CAB; ABC; BCA: Hence, the point P is the isogonal conjugate of the point Q wrt the triangle ABC (as long as the point P doesn lie on any of the lines BC; CA; AB; t since otherwise, the point Q is one of the vertices A; B; C of triangle ABC; and thus the isogonal conjugate of Q is not dened). Thus we have shown that if Q is the isogonal conjugate of a point P wrt a triangle ABC; then, in turn, the point P is the isogonal conjugate of the point Q wrt the triangle ABC (as long as the point P doesn lie on any of the lines BC; CA; AB). t Thus, instead of saying that "the point Q is the isogonal conjugate of the point P wrt the triangle ABC " or that "the point P is the isogonal conjugate of the point Q wrt the triangle ABC ", we can say that "the points P and Q are isogonally conjugate points wrt the triangle ABC ". Some rst properties of isogonal conjugates can be obtained by harvesting our proof of Theorem 3. We start with the following fact: Theorem 4. Let P and Q be two isogonally conjugate points wrt a triangle ABC: Then, the point P is an innite point if and only if the point Q lies on the circumcircle of triangle ABC: (See Fig. 10.)

11

Q B

C A

Fig. 10 Proof of Theorem 4. In order to show Theorem 4, we have to prove two assertions: Assertion 1: If the point P is an innite point, then the point Q lies on the circumcircle of triangle ABC: Assertion 2: If the point Q lies on the circumcircle of triangle ABC; then the point P is an innite point. Proof of Assertion 1. The point Q is the isogonal conjugate of P wrt triangle ABC; that is, the point of intersection of the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA: But since the point P is an innite point, this point of intersection must lie on the circumcircle of triangle ABC; as we saw in the proof of Theorem 3 in Case 2. Thus, Q lies on the circumcircle of triangle ABC; and Assertion 1 is proven. First proof of Assertion 2. Let P1 be the point of intersection of the line AP with the line at innity. Let Q1 be the isogonal conjugate of this point P1 wrt triangle ABC: Then, the line AQ1 is the isogonal of the line AP1 wrt the angle CAB: But since the point Q is the isogonal conjugate of the point P wrt triangle ABC; the line AQ is the isogonal of the line AP wrt the angle CAB: Since the lines AP1 and AP coincide, their isogonals wrt the angle CAB must also coincide; i. e., the lines AQ1 and AQ coincide. This means that the point Q1 lies on the line AQ: Since P1 is an innite 12

point, the already established Assertion 1 shows that its isogonal conjugate Q1 lies on the circumcircle of triangle ABC: Thus, the point Q1 is the point of intersection of the line AQ with the circumcircle of triangle ABC (dierent from A). But since the point Q lies on the circumcircle of triangle ABC; the point Q itself is the point of intersection of the line AQ with the circumcircle of triangle ABC (dierent from A). Hence, the points Q1 and Q coincide. Thus, the isogonal conjugates of these points Q1 and Q wrt triangle ABC must also coincide; but the isogonal conjugate of Q1 is the point P1 (since we have dened Q1 as the isogonal conjugate of P1 wrt triangle ABC), and the isogonal conjugate of Q is the point P: Hence, the points P1 and P coincide. Since P1 is an innite point, it follows that the point P is an innite point, and Assertion 2 is proven. Second proof of Assertion 2. We will show the following assertion, which is nothing but Assertion 2 with P and Q interchanged: Assertion 2 If the point P lies on the circumcircle of triangle ABC; then the point : Q is an innite point. This assertion 2 can be derived from our proof of Theorem 3 as follows: Since the point P lies on the circumcircle of triangle ABC; it is an Euclidean point; thus, according to the proof of Theorem 3 in Case 1, the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA are the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 : Since the point Q is the isogonal conjugate of the point P wrt triangle ABC; it is the point of intersection of these isogonals, thus the point of intersection of the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 : Now, as the point P lies on the circumcircle of triangle ABC; according to a well-known fact (Steiner line theorem), the re ections X 0 ; Y 0 ; Z 0 of the point P in the lines BC; CA; AB lie on one line; the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 are all perpendicular to this line X 0 Y 0 Z 0 and thus all parallel to each other. The point Q is the point of intersection of these parallel perpendicular bisectors, therefore an innite point (actually, it is the innite point of intersection of all lines perpendicular to the line X 0 Y 0 Z 0 and can be regarded as the "circumcenter" of the degenerate triangle X 0 Y 0 Z 0 ). Thus, Assertion 2 is proven, and therefore Assertion 2 as well. This completes our proof of Theorem 4. 4. Re ections and pedal circles Another almost trivial consequence of the proof of Theorem 3 is the following fact (Fig. 11): Theorem 5. If P is an Euclidean point in the plane of triangle ABC; and X 0 ; Y 0 ; Z 0 are the re ections of the point P in the lines BC; CA; AB; then the isogonal conjugate Q of the point P wrt the triangle ABC is the circumcenter of triangle X 0 Y 0 Z 0 ; and the lines AQ; BQ; CQ are the perpendicular bisectors of the sides Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 of this triangle. Proof of Theorem 5. Since P is an Euclidean point, we can retrieve from our proof of Theorem 3 in Case 1 the observation that the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA are the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 : But as the point Q is the isogonal conjugate of P wrt triangle ABC; the lines AQ; BQ; CQ are the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA: Hence, the lines AQ; BQ; CQ are the perpendicular bisectors of the 13

segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 : The point Q; being the point of intersection of the lines AQ; BQ; CQ; is therefore the point of intersection of the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 ; thus the circumcenter of triangle X 0 Y 0 Z 0 : This proves Theorem 5.

B X' Z'

P Q

C A

Y'
Fig. 11 Another easily accessible property of isogonals is the following one (Fig. 12): Theorem 6. If P is an Euclidean point in the plane of a triangle ABC; if X; Y; Z are the orthogonal projections of the point P on the lines BC; CA; AB; and if Q is the isogonal conjugate of the point P wrt the triangle ABC; then AQ ? Y Z; BQ ? ZX; CQ ? XY:

14

X Z P Q

Fig. 12 Proof of Theorem 6. Since the points Y and Z are the orthogonal projections of the point P on the lines CA and AB; according to Theorem 2 a) we see that the line Y Z is perpendicular to the isogonal of the line AP wrt the lines CA and AB; that is, to the isogonal of the line AP wrt the angle CAB: But as the point Q is the isogonal conjugate of P wrt triangle ABC; the line AQ is the isogonal of the line AP wrt the angle CAB: Hence, the line Y Z is perpendicular to the line AQ: Similarly, the lines ZX and XY are perpendicular to the lines BQ and CQ; and Theorem 6 is veried. We slowly move to deeper waters and show two well-known but less obvious properties of isogonal conjugates (Fig. 13): Theorem 7. Let P and Q be two Euclidean isogonally conjugate points wrt a triangle ABC: Let X; Y; Z be the orthogonal projections of the point P on the lines BC; CA; AB; and let U; V; W be the orthogonal projections of the point Q on the lines BC; CA; AB: a) We have P X QU = P Y QV = P Z QW: b) The points X; Y; Z; U; V; W lie on one circle centered at the midpoint M of the segment P Q: 15

U Z X

P W Q M

Fig. 13 Proof of Theorem 7. a) Since P is the isogonal conjugate of Q wrt the triangle ABC; the line AP is the isogonal of the line AQ wrt the angle CAB: Thus, ] (CA; AP ) = ] (AB; AQ) ; that is, ]Y AP = ]W AQ: Furthermore, ]AY P = ]AW Q (since ]AY P = 90 and ]AW Q = 90 ; and since we are working with directed angles modulo 180 ; we have 90 = 90 ). Thus, triangle AY P is oppositely similar to triangle AW Q: Similarly, triangle AZP is oppositely similar to triangle AV Q: Thus, the quadrilateral AY P Z; being formed by the triangles AY P and AZP; is oppositely similar to the quadrilateral AW QV; being formed by the triangles AW Q and AV Q: This similitude yields P Y : P Z = QW : QV; hence P Y QV = P Z QW: Analogous considerations lead to P X QU = P Y QV: Thus, P X QU = P Y QV = P Z QW; and Theorem 7 a) is proven. b) (See Fig. 14.) The point Q is the isogonal conjugate of P wrt the triangle ABC: Thus, according to Theorem 5, the point Q is the circumcenter of triangle X 0 Y 0 Z 0 ; where X 0 ; Y 0 ; Z 0 are the re ections of the point P in the lines BC; CA; AB: Hence, 0 0 0 QX = QY = QZ : Since X is the orthogonal projection of the point P on the line BC; while X 0 is the re ection of P in BC; we see that X is the midpoint of the segment P X 0 : On the other hand, M is the midpoint of the segment P Q: Thus, the segment M X is a midparallel QX 0 QY 0 QZ 0 in triangle P QX 0 ; so that M X = : Similarly, M Y = and M Z = : 2 2 2 0 0 0 Therefore, QX = QY = QZ entails M X = M Y = M Z: 16

As the conguration is symmetric in the points P and Q; we similarly obtain M U = M V = M W:

X' Z' X Z M Q P

Y'
Fig. 14 (See Fig. 15.) Since P X ? U X and QU ? U X; we have P X k QU ; thus, the quadrilateral P XU Q is a trapezoid with the bases P X and QU: Hence, the line through the midpoints of its legs P Q and U X is the midparallel of this trapezoid, therefore is parallel to its base P X; and thus, since P X ? U X; perpendicular to the line U X: Hence, this midparallel passes through the midpoint of the segment U X and is perpendicular to the line U X; thus, it is the perpendicular bisector of the segment U X: Since the midpoint of the segment P Q lies on this midparallel, we thus obtain that the midpoint of the segment P Q lies on the perpendicular bisector of the segment U X: The midpoint of the segment P Q is M ; thus, the point M lies on the perpendicular bisector of the segment U X: This yields M U = M X: Together with M X = M Y = M Z and M U = M V = M W; this yields M X = M Y = M Z = M U = M V = M W ; in other words, the points X; Y; Z; U; V; W lie on one circle centered at M: This proves Theorem 7 b).

17

P C

Fig. 15 Theorem 7 b) is known as the pedal circle theorem, and the proof we have given here is apparently new. 5. A result by Hatzipolakis, Yiu and Ehrmann The following observations are mostly due to Jean-Pierre Ehrmann ([6]) and, to smaller amounts, to Peter Scholze and me. We start with an interesting result of Ehrmann (Fig. 16): Theorem 8. Let P and Q be two Euclidean isogonally conjugate points wrt a triangle ABC: Then, the isogonal of the line AP wrt the angle BP C and the isogonal of the line AQ wrt the angle BQC are symmetric to each other wrt the line BC:

18

Fig. 16 Proof of Theorem 8. (See Fig. 17.) Let X 0 ; Y 0 ; Z 0 be the re ections of the point P ection of the point Q in the line BC: in the lines BC; CA; AB; and let U 0 be the re Then, the point Q is, in turn, the re ection of U 0 in the line BC: Since the points Q 0 and X are the re ections of the points U 0 and P in the line BC; the line QX 0 is the re ection of the line U 0 P (or, in other words, of the line P U 0 ) in the line BC: This means that the lines P U 0 and QX 0 are symmetric to each other wrt the line BC: According to Theorem 5, the point Q is the circumcenter of triangle X 0 Y 0 Z 0 ; thus the center of the circle through the points X 0 ; Y 0 ; Z 0 : Hence, by the central angle theorem for directed angles modulo 180 ; we have ]QX 0 Z 0 = 90 ]Z 0 Y 0 X 0 : In other words, ] (QX 0 ; Z 0 X 0 ) = 90 ] (Y 0 Z 0 ; X 0 Y 0 ) : Furthermore, Theorem 5 tells us that the lines AQ; BQ; CQ are the perpendicular bisectors of the segments Y 0 Z 0 ; Z 0 X 0 ; X 0 Y 0 ; this yields AQ ? Y 0 Z 0 ; BQ ? Z 0 X 0 ; CQ ? X 0 Y 0 ; and thus ] (Y 0 Z 0 ; AQ) = 90 ;

19

] (BQ; Z 0 X 0 ) = 90 and ] (X 0 Y 0 ; CQ) = 90 : Hence, ] (BQ; QX 0 ) = ] (BQ; Z 0 X 0 ) ] (QX 0 ; Z 0 X 0 ) = 90 (90 ] (Y 0 Z 0 ; X 0 Y 0 )) = ] (Y 0 Z 0 ; X 0 Y 0 ) = ] (Y 0 Z 0 ; AQ) ] (X 0 Y 0 ; AQ) = 90 ] (X 0 Y 0 ; AQ) = ] (X 0 Y 0 ; CQ) ] (X 0 Y 0 ; AQ) = ] (AQ; CQ) = ] (CQ; AQ) : Thus, the line QX 0 is the isogonal of the line AQ wrt the angle BQC: Similarly, the line P U 0 is the isogonal of the line AP wrt the angle BP C: Since we know that the lines P U 0 and QX 0 are symmetric to each other wrt the line BC; we have thus proven that the isogonal of the line AP wrt the angle BP C and the isogonal of the line AQ wrt the angle BQC are symmetric to each other wrt the line BC: Theorem 8 is thus established.

U' B X' Z'

C A

Y'
Fig. 17 The preceeding theorem is a crucial lemma in the proof of a result that was conjectured by Antreas P. Hatzipolakis, veried using computer algebra by Paul Yiu ([7]) and proven synthetically by Jean-Pierre Ehrmann ([6]): Theorem 9. Let P be a point in the plane of a triangle ABC: The lines AP; BP; CP intersect the lines BC; CA; AB at the points A0 ; B 0 ; C 0 : Let Q be the isogonal conjugate of the point P wrt the triangle ABC: Then, the re ections of the lines AQ; 20

BQ; CQ in the lines B 0 C 0 ; C 0 A0 ; A0 B 0 concur at one point. (See Fig. 18.)

A'

C' P

A B'

Fig. 18 Proof of Theorem 9. Again, we follow the proof given by Ehrmann in [6], rewriting it in a more elementary fashion. (See Fig. 19.) Let A1 ; B1 ; C1 ; P1 be the isogonal conjugates of the points A; B; C; P wrt the triangle A0 B 0 C 0 : Since P1 is the isogonal conjugate of P wrt triangle A0 B 0 C 0 ; the line A0 P1 is the isogonal of the line A0 P wrt the angle C 0 A0 B 0 : Since A1 is the isogonal conjugate of A wrt triangle A0 B 0 C 0 ; the line A0 A1 is the isogonal of the line A0 A wrt the angle C 0 A0 B 0 : Since the lines A0 P and A0 A coincide, their isogonals wrt the angle C 0 A0 B 0 must also coincide; i. e., the lines A0 P1 and A0 A1 coincide. Hence, the points A0 ; A1 ; P1 are collinear. Similarly, the points B 0 ; B1 and P1 are collinear, and the points C 0 ; C1 and P1 are collinear. Since B1 is the isogonal conjugate of B wrt triangle A0 B 0 C 0 ; the line A0 B1 is the isogonal of the line A0 B wrt the angle C 0 A0 B 0 : Since C1 is the isogonal conjugate of C wrt triangle A0 B 0 C 0 ; the line A0 C1 is the isogonal of the line A0 C wrt the angle C 0 A0 B 0 : Since the lines A0 B and A0 C coincide, their isogonals wrt the angle C 0 A0 B 0 must also 21

coincide; this means that the lines A0 B1 and A0 C1 coincide. In other words, the points A0 ; B1 ; C1 are collinear. Similarly, the points B 0 ; C1 ; A1 are collinear, and the points C 0 ; A1 ; B1 are collinear.

B1

A'

C' P P1 C1 A B' A1
Fig. 19 Since Q is the isogonal conjugate of the point P wrt triangle ABC; the line AQ is the isogonal of the line AP wrt the angle CAB: (See Fig. 20.) Since A and A1 are isogonally conjugate points wrt the triangle A0 B 0 C 0 ; Theorem 8 yields that the isogonal of the line A0 A wrt the angle B 0 AC 0 and the isogonal of the line A0 A1 wrt the angle B 0 A1 C 0 are symmetric to each other wrt the line B 0 C 0 : Now, the isogonal of the line A0 A wrt the angle B 0 AC 0 is the isogonal of the line AP wrt the angle CAB (since the line A0 A is the line A0 P; and the angle B 0 AC 0 is the angle CAB), and this is the line AQ: Further, the isogonal of the line A0 A1 wrt the angle B 0 A1 C 0 is the isogonal of the line A1 P1 wrt the angle C1 A1 B1 (since the line A0 A1 is the line A1 P1 ; and the angle B 0 A1 C 0 is the angle C1 A1 B1 ). Thus we obtained that the line AQ and the isogonal of the line A1 P1 wrt the angle C1 A1 B1 are symmetric to each other wrt the line B 0 C 0 : In other words, the re ection of the line AQ in the line 0 0 B C is the isogonal of the line A1 P1 wrt the angle C1 A1 B1 : Similarly, the re ections 22

of the lines BQ and CQ in the lines C 0 A0 and A0 B 0 are the isogonals of the lines B1 P1 and C1 P1 wrt the angles A1 B1 C1 and B1 C1 A1 : Altogether, the re ections of the lines AQ; BQ; CQ in the lines B 0 C 0 ; C 0 A0 ; A0 B 0 are the isogonals of the lines A1 P1 ; B1 P1 ; C1 P1 wrt the angles C1 A1 B1 ; A1 B1 C1 ; B1 C1 A1 ; and thus they concur at one point - at the isogonal conjugate of the point P1 wrt the triangle A1 B1 C1 : Theorem 9 is proven.

B1

A'

C' P P1 Q

C1 A A1
Fig. 20 6. Re ections in perpendicular bisectors As an application of isogonal conjugates, we are now going to prove some properties of the re ections of a point in the perpendicular bisectors of a triangle noted by Jos Carlos Chvez Sandoval in [5]. We start with some trivial facts: Theorem 10. Let P be an Euclidean point in the plane of a triangle ABC; and let D; E; F be the re ections of this point P in the perpendicular bisectors of the segments BC; CA; AB: a) The points P; D; E; F lie on one circle centered at the circumcenter O of triangle ABC: b) Triangle DEF is oppositely similar to triangle ABC: (See Fig. 21.) 23

C B'

O E P D A C

Fig. 21 Proof of Theorem 10. a) The circumcenter O of triangle ABC lies on the perpendicular bisector of its side BC: The point D is the re ection of the point P in this perpendicular bisector. Thus, OD = OP: Similarly, OE = OP and OF = OP: Consequently, OP = OD = OE = OF; so that the points P; D; E; F lie on one circle centered at O; and Theorem 10 a) is proven. b) As the points P; D; E; F lie on one circle, we have ]F DE = ]F P E: But as the point E is the re ection of P in the perpendicular bisector of CA; the line P E is perpendicular to the perpendicular bisector of CA: In turn, the perpendicular bisector of CA is perpendicular to the line CA: Hence, the line P E is parallel to the line CA: Similarly, the line P F is parallel to the line AB: Thus, ] (P F ; P E) = ] (AB; CA) ; what becomes ]F P E = ]BAC: Hence, ]F DE = ]F P E = ]BAC = ]CAB: Similarly, ]DEF = ]ABC: Consequently, the triangles DEF and ABC are oppositely similar, and Theorem 10 b) is proven. Now we come to a nontrivial property of triangle DEF: Before we formulate it we dene a traditional notion in triangle geometry: If S is the centroid of triangle ABC; and T is an arbitrary point in the plane, then 24

1 the image of the point T under the homothety with center S and factor is called 2 the complement of the point T wrt the triangle ABC: Now we show a theorem by Jos Carlos Chvez Sandoval ([5]): Theorem 11. Let P be an Euclidean point in the plane of a triangle ABC; and let D; E; F be the re ections of this point P in the perpendicular bisectors of the segments BC; CA; AB: Denote by AM ; BM ; CM the midpoints of the segments BC; CA; AB; and by DM ; EM ; FM the midpoints of the segments EF; F D; DE: Let Q be the isogonal conjugate of the point P wrt the triangle ABC; and let Q0 be the complement of the point Q wrt the triangle ABC: Then, the lines AM DM ; BM EM ; CM FM pass through the point Q0 : (See Fig. 22.)

Q CM S EM Q' E P A D
Fig. 22 Proof of Theorem 11. (See Fig. 23.) Since AM is the midpoint of the side BC of triangle ABC; the segment AAM is the A-median of this triangle; thus, it passes through the centroid S of triangle ABC and is divided by this centroid in the ratio AS 2 : 1: This means, we have = 2 (with directed segments). On the other hand, SAM the point Q0 is the complement of the point Q wrt triangle ABC; thus the image of the 25

DM

AM

BM FM

1 point Q under the homothety with center S and factor ; this means that the point 2 1 Q0 lies on the line SQ and satises SQ0 = SQ; so that 2 SQ0 = SQ = QS; and 2 AS AS QS QS = 2: Comparing this with = 2; we get = ; so that, by Thales, 0 SQ SAM SAM SQ0 AM Q0 k AQ:

AM

Q' S Q C

A
Fig. 23 (See Fig. 24.) Now, let Z 0 be the re ection of the point P in the line AB: Then, 0 ection wrt the perpendicular bisector ]Z AB = ]P AB: On the other hand, the re of the segment AB maps the points A and B to the points B and A; respectively, and the point P to the point F (because the point F was dened as the re ection of the point P in the perpendicular bisector of the segment AB). Since re ection wrt a line changes the sign of angles (but leaves them invariant in other respects), we thus have ]F BA = ]P AB: Together with ]Z 0 AB = ]P AB; this results in ]F BA = ]Z 0 AB; that means, ] (BF ; AB) = ] (AZ 0 ; AB) : Therefore, BF k AZ 0 : Similarly, AF k BZ 0 : Thus, the quadrilateral AF BZ 0 is a parallelogram. Using vectors, ! ! this rewrites as AZ 0 = F B: 26

Z' P C A
Fig. 24 (See Fig. 25.) After we have introduced the re ection Z 0 of the point P in the line AB; we also denote by Y 0 the re ection of the point P in the line CA: Similarly to ! ! ! ! 0 0 AZ = F B; we can then show AY = EC: Furthermore, according to Theorem 5, the line AQ is the perpendicular bisector of the segment Y 0 Z 0 ; consequently, the midpoint 0 XM of the segment Y 0 Z 0 lies on AQ: ! 0 0 On the other hand, since XM is the midpoint of the segment Y 0 Z 0 ; we have AXM = ! ! ! ! ! ! ! EC + F B ! ! AY 0 + AZ 0 0 0 0 : Since AY = EC and AZ = F B; this becomes AXM = : 2 2 Since AM is the midpoint of the segment BC; we have ! ! ! ! ! ! DM F + F B + EC EDM ! DM B + DM C DM A M = = : 2 2 ! ! But since DM is the midpoint of the segment EF; we have DM F = EDM ; so this simplies to ! ! ! ! ! ! EDM + F B + EC EDM ! ! EC + F B 0 = = AXM : DM AM = 2 2 0 0 Thus, DM AM k AXM : Since the line AXM coincides with the line AQ; we thus get DM AM k AQ: On the other hand, we know that AM Q0 k AQ: Therefore, DM AM k 27

AM Q0 : Now, the lines DM AM and AM Q0 have a common point (the point AM ), and thus can only be parallel if they coincide. As a consequence, the lines DM AM and AM Q0 coincide, i. e., the point Q0 lies on the line AM DM : Similarly, we may show that the point Q0 lies on the lines BM EM and CM FM ; and the proof of Theorem 11 is nished.

F AM Z' DM Q

X'M A

Y'
Fig. 25 7. More on isogonal conjugates and re ections The following theorem sheds light on a familiar conguration - re ections of a point in the sides of a triangle and its isogonal conjugate, see Theorem 5 - from a new, symmetric viewpoint:2 Theorem 12. Let A; B; C; D be four distinct points in the plane, and let A0 ; B 0 ; 0 C ; D0 be four distinct points in the plane. Then, the following four assertions B1 ; B2 ; B3 and B4 are pairwisely equivalent:
Of course, the point D in the following has nothing to do with the point D from Theorems 10 and 11, and the points A0 ; B 0 ; C 0 in the following have nothing to do with the points A0 ; B 0 ; C 0 from Theorem 9.
2

28

Assertion B1 : The points A; B; C; D are the circumcenters of triangles B 0 C 0 D0 ; C 0 D 0 A 0 ; D 0 A0 B 0 ; A 0 B 0 C 0 : Assertion B2 : The lines AB; BC; CD; DA; AC; BD are the perpendicular bisectors of the segments C 0 D0 ; D0 A0 ; A0 B 0 ; B 0 C 0 ; B 0 D0 ; A0 C 0 : Assertion B3 : The points A0 ; A0 ; A0 ; B 0 ; B 0 ; B 0 ; C 0 ; C 0 ; C 0 ; D0 ; D0 ; D0 are the re ections of the points B 0 ; C 0 ; D0 ; C 0 ; D0 ; A0 ; D0 ; A0 ; B 0 ; A0 ; B 0 ; C 0 in the lines CD; DB; BC; DA; AC; CD; AB; BD; DA; BC; CA; AB: Assertion B4 : The points A0 ; B 0 ; C 0 ; D0 are the isogonal conjugates of the points A; B; C; D wrt the triangles BCD; CDA; DAB; ABC: (See Fig. 26.)

A'

D C'

D'

B'
Fig. 26 Proof of Theorem 12. First, we show the equivalence of the assertions B1 and B2 : If Assertion B1 holds, then the points A; B; C; D are the circumcenters of triangles 0 0 0 B C D ; C 0 D0 A0 ; D0 A0 B 0 ; A0 B 0 C 0 : Then, since the circumcenter of a triangle lies on the perpendicular bisectors of its sides, the point A; being the circumcenter of triangle B 0 C 0 D0 ; must lie on the perpendicular bisector of the segment C 0 D0 : Similarly, the point B must lie on the perpendicular bisector of the segment C 0 D0 : Hence, the line AB is the perpendicular bisector of the segment C 0 D0 : Similarly, the lines BC; CD; 29

DA; AC; BD are the perpendicular bisectors of the segments D0 A0 ; A0 B 0 ; B 0 C 0 ; B 0 D0 ; A0 C 0 ; and thus Assertion B2 is fullled. Conversely, if Assertion B2 holds, then the lines AB; BC; CD; DA; AC; BD are the perpendicular bisectors of the segments C 0 D0 ; D0 A0 ; A0 B 0 ; B 0 C 0 ; B 0 D0 ; A0 C 0 : Hence, the point A; being the point of intersection of the lines AB; AC; DA; must be the point of intersection of the perpendicular bisectors of the segments C 0 D0 ; B 0 D0 ; B 0 C 0 ; and thus the circumcenter of the triangle B 0 C 0 D0 : Similarly, the points B; C; D are the circumcenters of the triangles C 0 D0 A0 ; D0 A0 B 0 ; A0 B 0 C 0 : Thus, Assertion B1 must hold. Hence, we have shown that the assertions B1 and B2 are equivalent. The equivalence of the assertions B3 and B2 evidently follows from the following obvious fact: If P1 and P2 are two distinct points and g1 is a line, then the point P2 is the re ection of the point P1 in the line g1 if and only if the line g1 is the perpendicular bisector of the segment P1 P2 : Altogether we have now shown that the assertions B1 ; B2 and B3 are pairwisely equivalent. In order to verify the equivalence of all four assertions B1 ; B2 ; B3 and B4 ; it remains to prove that the assertions B3 and B4 are equivalent. In order to prove this, we have to establish two auxiliary results: Auxiliary result 1. If Assertion B4 holds, then so does Assertion B3 : Auxiliary result 2. If Assertion B3 holds, then so does Assertion B4 : Proof of Auxiliary result 1. Assume that Assertion B4 holds. Then, particularly, the point B 0 is the isogonal conjugate of the point B wrt triangle CDA: Thus, the line AB 0 is the isogonal of the line AB wrt the angle DAC: Hence, ] (AC; AB 0 ) = ] (DA; AB) : Equivalently, ]CAB 0 = ]DAB: Similarly, ]CAD0 = ]BAD: 0 ection Thus, ]CAB 0 = ]DAB = ( ]BAD) = ]CAD0 : Now, if B1 is the re 0 0 0 0 of the point D in the line AC; then ]CAB1 = ]CAD ; so that ]CAB1 = ]CAB 0 : 0 0 Hence, the point B1 lies on the line AB 0 : Similarly, the point B1 lies on the line CB 0 : But the lines AB 0 and CB 0 have only one point in common, namely the point B 0 : 0 0 Thus, since the point B1 lies on both of these lines, we must have B1 = B 0 : Since we 0 0 ection of the point D in the line AC; we thus have introduced the point B1 as the re 0 conclude that the point B is the re ection of the point D0 in the line AC: Similarly, 0 0 0 0 0 0 0 the points A ; A ; A ; B ; B ; C ; C ; C 0 ; D0 ; D0 ; D0 are the re ections of the points B 0 ; C 0 ; D0 ; C 0 ; A0 ; D0 ; A0 ; B 0 ; A0 ; B 0 ; C 0 in the lines CD; DB; BC; DA; CD; AB; BD; DA; BC; CA; AB: In other words, Assertion B3 holds. This proves Auxiliary result 1. First proof of Auxiliary result 2. Assume that Assertion B3 is valid. Since the assertions B1 ; B2 and B3 are equivalent, it thus follows that Assertions B1 and B2 hold as well; i. e., the points A; B; C; D are the circumcenters of triangles B 0 C 0 D0 ; C 0 D0 A0 ; D0 A0 B 0 ; A0 B 0 C 0 ; and the lines AB; BC; CD; DA; AC; BD are the perpendicular bisectors of the segments C 0 D0 ; D0 A0 ; A0 B 0 ; B 0 C 0 ; B 0 D0 ; A0 C 0 : Since the lines DA; AB; AC are the perpendicular bisectors of the segments B 0 C 0 ; C 0 D0 ; B 0 D0 ; we have DA ? B 0 C 0 ; AB ? C 0 D0 ; AC ? B 0 D0 ; so that ] (DA; B 0 C 0 ) = 90 ; ] (AB; C 0 D0 ) = 90 ; ] (B 0 D0 ; AC) = 90 : Thus, ] (DA; AB) = ] (DA; B 0 C 0 ) + ] (B 0 C 0 ; AB) = 90 + ] (B 0 C 0 ; AB) = ] (AB; C 0 D0 ) + ] (B 0 C 0 ; AB) = ] (B 0 C 0 ; C 0 D0 ) = ]B 0 C 0 D0 : On the other hand, since A is the circumcenter of triangle B 0 C 0 D0 ; thus the center of a circle through the points B 0 ; C 0 ; D0 ; the central angle theorem yields ]D0 B 0 A = 30

90

]B 0 C 0 D0 : Hence, ] (AC; AB 0 ) = ] (B 0 D0 ; AB 0 ) ] (B 0 D0 ; AC) = ]D0 B 0 A = (90 ]B 0 C 0 D0 ) 90 = ]B 0 C 0 D0 : 90

Comparing this with ] (DA; AB) = ]B 0 C 0 D0 ; we infer ] (AC; AB 0 ) = ] (DA; AB) : Thus, the line AB 0 is the isogonal of the line AB wrt the angle DAC: Similarly, the lines CB 0 and DB 0 are the isogonals of the lines CB and DB wrt the angles ACD and CDA: Thus, the point B 0 is the point of intersection of the isogonals of the lines CB; DB; AB wrt the angles ACD; CDA; DAC: In other words, the point B 0 is the isogonal conjugate of the point B wrt the triangle CDA: Similarly, the points C 0 ; D0 ; A0 are the isogonal conjugates of the points C; D; A wrt the triangles DAB; ABC; BCD: Hence, Assertion B4 holds; this proves Auxiliary result 2. Second proof of Auxiliary result 2. It is particularly easy to prove Auxiliary result 2 basing on Theorem 5: Assume that Assertion B3 holds. Then, in particular, the point A0 is the re ection 0 0 0 0 0 of the point B in the line CD: Hence, DA = DB : Similarly, DB = DC : Thus, DA0 = DB 0 = DC 0 ; and this signies that the point D is the circumcenter of triangle A0 B 0 C 0 : Since we assumed Assertion B3 to hold, the points A0 ; B 0 ; C 0 are the re ections of 0 the point D in the lines BC; CA; AB: Thus, after Theorem 5, the isogonal conjugate of the point D0 wrt triangle ABC is the circumcenter of triangle A0 B 0 C 0 : But as we know that the point D is the circumcenter of triangle A0 B 0 C 0 ; it follows that the point D is the isogonal conjugate of the point D0 wrt triangle ABC: Hence, in turn, the point D0 is the isogonal conjugate of the point D wrt triangle ABC: Similarly, the points A0 ; B 0 ; C 0 are the isogonal conjugates of the points A; B; C wrt triangles BCD; CDA; DAB: Therefore, Assertion B4 holds. This again proves Auxiliary result 2. This completes the proof of Theorem 12. An easy consequence of Theorem 12 is (Fig. 27): Theorem 13. Let P be an Euclidean point in the plane of triangle ABC; and let 0 X ; Y 0 ; Z 0 be the re ections of this point P in the lines BC; CA; AB: Let Q be the isogonal conjugate of the point P wrt triangle ABC: Then, the points X 0 ; Y 0 ; Z 0 are the isogonal conjugates of the points A; B; C wrt the triangles BQC; CQA; AQB:

31

X' Z'

P Q

C A

Y'
Fig. 27 Proof of Theorem 13. (See Fig. 28.) Since Y 0 is the re ection of the point P in 0 0 the line CA; we have AY = AP: Similarly, AZ = AP: Therefore, AY 0 = AP = AZ 0 ; thus, the point A is the circumcenter of triangle Y 0 Z 0 P: Similarly, BZ 0 = BP = BX 0 and CX 0 = CP = CY 0 ; what yields that the points B and C are the circumcenters of triangles Z 0 P X 0 and P X 0 Y 0 : Further, according to Theorem 5, the point Q is the circumcenter of triangle X 0 Y 0 Z 0 : Altogether, we see that the points A; B; C; Q are the circumcenters of triangles Y 0 Z 0 P; Z 0 P X 0 ; P X 0 Y 0 ; X 0 Y 0 Z 0 : Thus, the four distinct points A; B; C; Q and the four distinct points X 0 ; Y 0 ; Z 0 ; P fulll the Assertion B1 of Theorem 12. But since, according to Theorem 12, the Assertion B1 is equivalent to Assertion B4 ; these points must therefore also satisfy Assertion B4 : In other words, the points X 0 ; Y 0 ; Z 0 ; P are the isogonal conjugates of the points A; B; C; Q wrt the triangles BCQ; CQA; QAB; ABC: Equivalently, the points X 0 ; Y 0 ; Z 0 ; P are the isogonal conjugates of the points A; B; C; Q wrt the triangles BQC; CQA; AQB; ABC: This implies Theorem 13.

32

B X' Z'

C A

Y'
Fig. 28 Theorem 13 simplies our proof of Theorem 8 given above. In fact, according to Theorem 13, the point X 0 is the isogonal conjugate of the point A wrt triangle BQC; what immediately yields that the line QX 0 is the isogonal of the line QA (that is, the line AQ) wrt the angle BQC; this was a crucial result in the proof of Theorem 8. [Thanks to Marcello Tarquini for reminding me of this shortcut.] 8. Isogonal conjugates of basic centers It is useful to identify the isogonal conjugates of known triangle centers. We start with a trivial fact: Theorem 14. Let P be a point in the plane of a triangle ABC: Then, the isogonal conjugate of the point P wrt triangle ABC coincides with the point P if and only if the point P is the incenter or one of the three excenters of triangle ABC: (See Fig. 29.)

33

C A

Fig. 29 Proof of Theorem 14. According to the denition of isogonals, the isogonal of the line AP wrt the angle CAB coincides with the line AP if and only if ] (AB; AP ) = ] (CA; AP ) : But this is equivalent to the line AP being either the internal or the external angle bisector of the angle CAB: Hence, the isogonal of the line AP wrt the angle CAB coincides with the line AP if and only if the line AP is either the internal or the external angle bisector of the angle CAB: Similarly, the isogonal of the line BP wrt the angle ABC coincides with the line BP if and only if the line BP is either the internal or the external angle bisector of the angle ABC; and the isogonal of the line CP wrt the angle BCA coincides with the line CP if and only if the line CP is either the internal or the external angle bisector of the angle BCA: The isogonal conjugate of the point P wrt triangle ABC is the point of intersection of the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA: Consequently, the isogonal conjugate of the point P wrt triangle ABC coincides with the point P if and only if the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA intersect at the point P; i. e. if and only if the isogonals of the lines AP; BP; CP wrt the angles CAB; ABC; BCA coincide with the lines AP; BP; CP: As we know, this is equivalent to the lines AP; BP; CP being the internal or external angle 34

bisectors of the angles CAB; ABC; BCA: But there are exactly four points P in the plane of triangle ABC such that the lines AP; BP; CP are the internal or external angle bisectors of the angles CAB; ABC; BCA; and these points are the incenter and the three excenters of triangle ABC: Thus it is proven that the isogonal conjugate of the point P wrt triangle ABC coincides with the point P if and only if the point P is the incenter or one of the three excenters of triangle ABC: This proves Theorem 14. A slightly less trivial fact is the following (Fig. 30): Theorem 15. Let O be the circumcenter and H the orthocenter of a triangle ABC: Then, the points O and H are isogonally conjugate points wrt triangle ABC: In brief: The circumcenter and the orthocenter of any triangle are isogonally conjugate points wrt this triangle. First proof of Theorem 15. Being the circumcenter of triangle ABC; the point O is the center of a circle through the points A; B; C: Thus, according to the central angle theorem, ]CAO = 90 ]ABC: On the other hand, since H is the orthocenter of triangle ABC; we know that AH ? BC; so that ] (AH; BC) = 90 ; and therefore ] (AB; AH) = ] (AB; BC) ] (AH; BC) = ]ABC = ]CAO = ] (CA; AO) : 90 = (90 ]ABC)

Hence, the line AH is the isogonal of the line AO wrt the angle CAB: Similarly, the lines BH and CH are the isogonals of the lines BO and CO wrt the angles ABC and BCA: Thus, the point H; being the point of intersection of these lines AH; BH; CH; must be the isogonal conjugate of the point O wrt triangle ABC: This proves Theorem 15.

35

O H

C A

Fig. 30 Second proof of Theorem 15. (See Fig. 31.) Let BM and CM be the midpoints of the sides CA and AB of triangle ABC: Then, BM CM k BC: On the other hand, since the circumcenter O of triangle ABC lies on the perpendicular bisectors of its sides CA and AB; the midpoints BM and CM of these sides CA and AB must be the orthogonal projections of the point O on the lines CA and AB: Hence, by Theorem 2 a) (applied to the two lines CA and AB and the point O in the plane), the line BM CM is perpendicular to the isogonal of the line AO wrt the lines CA and AB; i. e. wrt the angle CAB: In other words: The isogonal of the line AO wrt the angle CAB is perpendicular to the line BM CM : Since BM CM k BC; this isogonal must therefore be perpendicular to the line BC; and since this isogonal passes through the point A; we can conclude that it is the A-altitude of triangle ABC and thus passes through its orthocenter H: So the point H lies on the isogonal of the line AO wrt the angle CAB: Similarly, the point H lies on the isogonals of the lines BO and CO wrt the angles ABC and BCA: Thus, the point H is the point of intersection of these isogonals, i. e. the isogonal conjugate of the point O wrt triangle ABC: Once again Theorem 15 is proven. 36

CM O

H C A BM

Fig. 31 At this point, we mention an easy consequence of Theorem 15 which will be of use to us later (Fig. 32): Theorem 16. Let AH ; BH ; CH be the feet of the altitudes of triangle ABC from the vertices A; B; C: Let O be the circumcenter of triangle ABC: Then, AO ? BH CH ; BO ? CH AH ; CO ? AH BH : Proof of Theorem 16. Obviously, the orthocenter H of triangle ABC lies on the altitude of triangle ABC from the vertex A: The point AH is the foot of this altitude. Thus, the point AH is the orthogonal projection of the point H on the line BC: Similarly, the points BH and CH are the orthogonal projections of the point H on the lines CA and AB: By Theorem 15, the point O is the isogonal conjugate of the point H wrt triangle ABC: Thus, applying Theorem 6 to the point H in the plane of triangle ABC; we get AO ? BH CH ; BO ? CH AH ; CO ? AH BH : This proves Theorem 16.

37

AH

CH O H

C A BH

Fig. 32 Theorem 16 can also be directly shown through an angle chase. 9. Symmedians and antiparallels (See Fig. 33.) Let S be the centroid of triangle ABC: Then, the lines AS; BS; and CS are the A-median, the B-median, and the C-median of triangle ABC; respectively. The isogonals of these medians AS; BS; CS wrt the angles CAB; ABC; BCA are called the A-symmedian, B-symmedian, C-symmedian of triangle ABC; respectively, and altogether referred to as the three symmedians of triangle ABC: As a consequence of their denition, these symmedians of triangle ABC intersect at one point, namely at the isogonal conjugate of the point S wrt triangle ABC: This isogonal conjugate of the point S wrt triangle ABC is called the symmedian point of triangle ABC and will be denoted by K in the following. Then, as the A-symmedian, the B-symmedian, and the C-symmedian of triangle ABC intersect at the point K; they are the lines AK; BK; and CK: In brief: The symmedian point of a triangle is the point of intersection of its symmedians, and it is the isogonal conjugate of the centroid of this triangle wrt this triangle. 38

S K

C A

Fig. 33 We won discuss further properties of the symmedian point here, but we use the t occasion to give an introduction into the often utilized concept of antiparallels wrt a side of a triangle: Let ABC be a triangle, and g a line in the plane. The line g is said to be antiparallel to BC wrt the triangle ABC if and only if it is parallel to the tangent to the circumcircle of triangle ABC at the point A: In this case, one also uses to say that the line g is an antiparallel to the side BC of triangle ABC: Similarly, we dene when a line is called antiparallel to CA or to AB wrt triangle ABC; or, in other words, when it is an antiparallel to the side CA or to the side AB of triangle ABC: An important remark about this notion of antiparallelism is that the meaning of the word "antiparallel", in contrast to the meaning of the word "parallel", depends on the triangle ABC: Just to say that two lines are antiparallel to each other doesn t make sense; one can only say that a line is antiparallel to a side of a triangle wrt this triangle. Accordingly, in the formulation "the line g is an antiparallel to the side BC of triangle ABC ", one cannot omit the clause "of triangle ABC ", since it

39

species the triangle of reference for the notion "antiparallel". Clearly, through any point there exists exactly one antiparallel to the side BC of triangle ABC (in fact, by its denition, an antiparallel to the side BC of triangle ABC means a parallel to the tangent to the circumcircle of triangle ABC at the point A; and through any point there exists exactly one parallel to this tangent). Similarly, through any point there exists exactly one antiparallel to the side CA of triangle ABC and exactly one antiparallel to the side AB of triangle ABC: The basic advantage of the notion of antiparallelism are the many equivalent criteria for a line to be antiparallel to a side of a triangle. Some of these criteria are given by the following theorem3 : Theorem 17. Let ABC be a triangle, and let g be a line in the plane. Consider the following ten assertions: Assertion E0 : The line g is antiparallel to BC wrt triangle ABC: Assertion E1 : The line g is parallel to the tangent to the circumcircle of triangle ABC at the point A: Assertion E2 : The line g is perpendicular to the line AO; where O is the circumcenter of triangle ABC: Assertion E3 : We have ] (CA; g) = ] (BC; AB) : Assertion E4 : We have ] (AB; g) = ] (BC; CA) : Assertion E5 : The line g is parallel to the line BH CH ; where BH and CH are the feet of the B-altitude and the C-altitude of triangle ABC: 4 Assertion E6 : If P and Q are the points of intersection of the line g with the lines CA and AB; then the triangles AP Q and ABC are oppositely similar. Assertion E7 : If P and Q are the points of intersection of the line g with the lines CA and AB; then there exists a circle which meets the line CA at the points C and P and meets the line AB at the points B and Q: 5 Assertion E8 : If P and Q are the points of intersection of the line g with the lines CA and AB; then AB AQ = AC AP; where the segments are directed. Assertion E9 : If P and Q are the points of intersection of the line g with the lines CA and AB; then the midpoint of the segment P Q lies on the A-symmedian of triangle
Of course, the points P and Q from Theorem 17 have nothing to do with the points P and Q from the above results on isogonal conjugates. 4 If triangle ABC is right-angled at A; then these feet BH and CH both coincide with the vertex A; in this case, the line BH CH is to be understood as the tangent to the circumcircle of triangle ABC at the point A: 5 Hereby, the following convention applies: If a circle k touches a line g at a point T; then we say that the circle k meets the line g at the points T and T: This convention claries how the Assertion E7 is to be understood if e. g. the point P coincides with the point C: In this case, Assertion E7 states that there exists a circle which meets the line CA at the points C and C (that is, touches the line CA at the point C) and meets the line AB at the points B and Q: This is the reason why the formulation "there exists a circle which meets the line CA at the points C and P and meets the line AB at the points B and Q " is superior to the shorter formulation "the points B; C; P; Q lie on one circle". In fact, if P 6= C and Q 6= B; these formulations are equivalent, but e. g. in the case when the point P coincides with the point C; the points B; C; P; Q always lie on one circle; hence, if we had used the formulation "the points B; C; P; Q lie on one circle" for Assertion E7 ; then this Assertion E7 would not be equivalent to E0 in the case P = C (and similarly in the case Q = B as well).
3

40

ABC: Then, we have: a) The six assertions E0 ; E1 ; E2 ; E3 ; E4 and E5 are pairwisely equivalent. b) If the line g doesn pass through the point A; then the ten assertions E0 ; E1 ; E2 ; t E3 ; E4 ; E5 ; E6 ; E7 ; E8 and E9 are pairwisely equivalent. (See Fig. 34 for Assertions E1 ; E2 ; E3 ; E4 ; Fig. 36 for Assertion E5 ; Fig. 37 for Assertions E6 ; E7 ; E8 ; and Fig. 38 for Assertion E9 :) Theorem 17 yields, altogether, nine criteria for a line to be antiparallel to BC wrt triangle ABC: Similar criteria hold for antiparallelism to CA or to AB:

g tA

C A

Fig. 34 Our proof of Theorem 17 will require a fact which slightly extends the intersecting chords theorem, the intersecting secants theorem, and the intersecting secant and tangent theorem:

41

U'

V' P v V

Fig. 35 Theorem 18. Let u and v be two lines which intersect at an Euclidean point P: Let U and U 0 be two points on the line u distinct from P; and let V and V 0 be two points on the line v distinct from P: Then, the following three assertions D1 ; D2 and D3 are pairwisely equivalent: Assertion D1 : The triangles P U V and P V 0 U 0 are oppositely similar. Assertion D2 : There exists a circle which meets the line u at the points U and U 0 and meets the line v at the points V and V 0 : 6 Assertion D3 : We have P U P U 0 = P V P V 0 ; where the segments are directed. (See Fig. 35.) Proof of Theorem 18.7 We consider only the case when U 6= U 0 and V 6= V 0 : The
Hereby, the same convention applies as in Assertion E7 of Theorem 17. This proof is given here but for the sake of consequence in our application of directed angles modulo 180 : In fact, the equivalence of Assertions D1 and D3 is a trivial corollary of a well-known similitude criterion which states that two triangles P1 P2 P3 and Q1 Q2 Q3 are oppositely similar if and only if \P1 P2 P3 = \Q1 Q2 Q3 und P1 P2 : P2 P3 = Q1 Q2 : Q2 Q3 : But here, the sign \ stands for directed angles modulo 360 ; such a similitude criterion cannot hold for directed angles modulo 180 : As we aim at using directed angles modulo 180 throughout this paper, we will prove this equivalence in a dierent way.
7 6

42

cases U = U 0 and V = V 0 can be handled in the same way, with the only dierence that here and there instead of chordal angles, we have angles between a chord and a tangent. First we show the equivalence of the Assertions D1 and D2 : Assertion D1 states that triangles P U V and P V 0 U 0 are oppositely similar. This is equivalent to ]P U V = ]P V 0 U 0 (in fact, if triangles P U V and P V 0 U 0 are oppositely similar, then ]P U V = ]P V 0 U 0 ; and conversely, if ]P U V = ]P V 0 U 0 ; then, together with ]U P V = ]V 0 P U 0 ; this yields the opposite similarity of triangles P U V and P V 0 U 0 ). But ]P U V = ]P V 0 U 0 rewrites as ]U 0 U V = ]U 0 V 0 V; and this equation holds if and only if the points U; V; U 0 and V 0 lie on one circle. Since U 6= U 0 and V 6= V 0 ; the points U; V; U 0 and V 0 lie on one circle if and only if there exists a circle which meets the line u at the points U and U 0 and meets the line v at the points V and V 0 : The latter is Assertion D2 : Combining these steps, we realize that we have proven the equivalence of Assertions D1 and D2 : Now we are going to prove the equivalence of Assertions D2 and D3 : If Assertion D2 holds, then there exists a circle which meets the line u at the points U and U 0 and meets the line v at the points V and V 0 : The power of the point P wrt this circle equals P U P U 0 on the one hand, and equals P V P V 0 on the other hand. Thus, P U P U 0 = P V P V 0 ; so that Assertion D3 is valid. Conversely: If we assume Assertion D3 to hold, then P U P U 0 = P V P V 0 : Now, let V10 be the point of intersection of the circle through the points U; U 0 and V with the line v dierent from V: 8 Then, this circle meets the line u at the points U and U 0 and meets the line v at the points V and V10 : Consequently, the power of the point P wrt this circle equals P U P U 0 on the one hand, and equals P V P V10 on the other hand. Thus, P U P U 0 = P V P V10 : Comparison with P U P U 0 = P V P V 0 yields P V P V10 = P V P V 0 ; thus P V10 = P V 0 (since P V 6= 0). Since the points V10 and V 0 both lie on the line v and since we use directed segments, this entails that the points V10 and V 0 coincide. The fact that the circle through the points U; U 0 and V meets the line u at the points U and U 0 and meets the line v at the points V and V10 can now be rewritten as follows: The circle through the points U; U 0 and V meets the line u at the points U and U 0 and meets the line v at the points V and V 0 : Thus, Assertion D2 is fullled. Hence, we have shown the equivalence of Assertions D2 and D3 : Altogether, we subsume that all three Assertions D1 ; D2 and D3 are equivalent, and Theorem 18 is proven. Proof of Theorem 17. a) The equivalence of Assertions E0 and E1 is a paraphrase of the denition of antiparallelism. In order to show the equivalence of Assertions E1 and E2 ; it is enough to show that the tangent to the circumcircle of triangle ABC at the point A is perpendicular to the line AO: But this is clear, since O is the center of this circumcircle.
8

If this circle happens to touch the line v; then we set V10 = V:

43

CH

g O

C A BH

Fig. 36 The equivalence of Assertions E1 and E3 can be proven as follows: (See Fig. 35.) Let tA be the tangent to the circumcircle of triangle ABC at the point A: Then, Assertion E1 states that g k tA : This is equivalent to ] (CA; g) = ] (CA; tA ) : But since tA is the tangent to the circumcircle of triangle ABC at the point A; while ]CBA is the chordal angle of the chord CA in this circumcircle, we have ] (CA; tA ) = ]CBA; what rewrites as ] (CA; tA ) = ] (BC; AB) : Hence, the equation ] (CA; g) = ] (CA; tA ) is equivalent to the equation ] (CA; g) = ] (BC; AB) : But this equation is Assertion E3 : Hence it is shown that Assertion E1 is equivalent to Assertion E3 : An analogous argument shows the equivalence of the Assertions E1 and E4 : In order to prove the equivalence of the Assertions E2 and E5 ; it is obviously su cient to verify that AO ? BH CH : But this follows from Theorem 16.9 Altogether, we have proven all six Assertions E0 ; E1 ; E2 ; E3 ; E4 and E5 to be equivalent. Thus, the proof of Theorem 17 a) is complete.
The relation AO ? BH CH also holds in the case when triangle ABC is right-angled at A: In fact, in this case we have specied that the line BH CH is the tangent to the circumcircle of triangle ABC at the point A; and this tangent is perpendicular to the line AO (as we already saw).
9

44

C A P

Fig. 37 b) As Theorem 17 a) is already demonstrated, we know that the Assertions E0 ; E1 ; E2 ; E3 ; E4 and E5 are all equivalent. It just remains to prove the equivalence of the Assertions E6 ; E7 ; E8 and E9 to these Assertions in the case when the line g doesn pass t through the point A: First we establish the equivalence of Assertions E3 and E6 : If Assertion E3 holds, then ] (CA; g) = ] (BC; AB) ; this rewrites as ]AP Q = ]ABC: On the other hand, ]P AQ = ]BAC: Hence, the triangles AP Q and ABC are oppositely similar, so that Assertion E6 is valid. Conversely: If Assertion E6 holds, then triangles AP Q and ABC are oppositely similar, what yields ]AP Q = ]ABC: In other words, ] (CA; g) = ] (BC; AB) : Hence, Assertion E3 is valid. Thus we have shown the equivalence of the Assertions E3 and E6 : The equivalence of the Assertions E6 ; E7 and E8 follows from Theorem 18, applied to the two lines CA and AB which intersect at the Euclidean point A; the two points P and C on the line CA distinct from A; and the two points Q and B on the line AB distinct from A: 45

g Q AM

K M0

C A P

Fig. 38 Altogether, we now know that the nine Assertions E0 ; E1 ; E2 ; E3 ; E4 ; E5 ; E6 ; E7 and E8 are all pairwisely equivalent. In order to prove the equivalence of all ten Assertions E0 ; E1 ; E2 ; E3 ; E4 ; E5 ; E6 ; E7 ; E8 and E9 ; it therefore su ces to prove the equivalence of the Assertions E0 and E9 : This proof is obtained by showing the following two auxiliary results: Auxiliary result 1. If Assertion E0 holds, then Assertion E9 holds. Auxiliary result 2. If Assertion E9 holds, then Assertion E0 holds. Proof of Auxiliary result 1. (See Fig. 38.) Assume that Assertion E0 holds. As we already know that Assertion E0 is equivalent to Assertion E6 ; it thus follows that Assertion E6 holds as well. That is, the triangles AP Q and ABC are oppositely similar. In oppositely similar triangles, corresponding points form oppositely equal angles. Let M0 be the midpoint of the segment P Q; and AM the midpoint of the segment BC: Then, the points M0 and AM are corresponding points in the oppositely similar triangles AP Q and ABC (being the midpoints of their respective sides P Q and BC). Hence, they form oppositely equal angles; particularly, ]P AM0 = ]BAAM : In other words, ] (CA; AM0 ) = ] (AB; AAM ) : Hence, the line AM0 is the isogonal of the line AAM 46

wrt the angle CAB: But since AM is the midpoint of the segment BC; the line AAM is the A-median of triangle ABC: Therefore, the line AM0 is the isogonal of the Amedian of triangle ABC wrt the angle CAB; hence the A-symmedian of triangle ABC: Consequently, the point M0 lies on the A-symmedian of triangle ABC: Since M0 is the midpoint of the segment P Q; this is exactly Assertion E9 : Thus, Auxiliary result 1 is proven. Proof of Auxiliary result 2. (See Fig. 39. This gure is intentionally drawn wrong in order not to tempt to unfounded conclusions.) Assume that Assertion E9 is valid. In other words, the midpoint M0 of the segment P Q lies on the A-symmedian of triangle ABC: Let the antiparallel to the side BC of triangle ABC through the point M0 intersect the lines CA and AB at the points P 0 and Q0 : Since the line P 0 Q0 is antiparallel to BC wrt triangle ABC; it fullls Assertion E0 of Theorem 17. As we have already shown that Assertion E0 implies Assertion E9 (this was our Auxiliary result 1), we thus conclude that this line P 0 Q0 fullls Assertion E9 : That is: The midpoint of the segment P 0 Q0 lies on the A-symmedian of triangle ABC: Hence, the midpoint of the segment P 0 Q0 is the point of intersection of the line P 0 Q0 with the A-symmedian of triangle ABC: But this point of intersection is the point M0 : Therefore, the midpoint of the segment P 0 Q0 must be the point M0 : Since M0 is the midpoint of the segment P Q; the re ection with respect to the point M0 maps the point P to the point Q: Since M0 is the midpoint of the segment P 0 Q0 ; the re ection with respect to the point M0 maps the point P 0 to the point Q0 : Now, if the points P and P 0 were distinct, then the points Q and Q0 would therefore be distinct as well (since a re ection maps distinct points to distinct points), and we 0 0 would have QQ k P P (since a re ection maps a line to a parallel line); but this cannot 0 be true, since the lines QQ and P P 0 are the lines AB and CA; and we have AB , CA: Hence, the points P and P 0 cannot be distinct, i. e. we must have P = P 0 : Similarly, Q = Q0 : As we know that the line P 0 Q0 is antiparallel to BC wrt triangle ABC; we can thus conclude that the line P Q is antiparallel to BC wrt triangle ABC: This means that Assertion E0 holds. Thus, Auxiliary result 2 is proven. This completes the proof of Theorem 17.

47

Q Q' g K M0

P' A P

Fig. 39 The notion of antiparallelism is mostly applied in the theory of Tucker circles, but can also often be used to simplify some standard ways of conclusion in elementary geometry. 10. Isogonal conjugates of Kiepert points A means for identication of isogonal conjugates is the following fact (Fig. 40): Theorem 19. Let ABC be a triangle, and let P and Q be two points on the perpendicular bisector of the segment BC: a) The following ve assertions F1 ; F2 ; F3 ; F4 and F5 are pairwisely equivalent: Assertion F1 : We have ]ABP = ]ACQ: Assertion F2 : We have ]ACP = ]ABQ: Assertion F3 : We have ]BCP + ]BCQ = ]BAC: Assertion F4 : We have ]P BC + ]QBC = ]BAC: Assertion F5 : The points P and Q are inverse to each other wrt the circumcircle of triangle ABC: 48

C A

Fig. 40 b) If one of the ve Assertions F1 ; F2 ; F3 ; F4 and F5 holds, then the lines AP and AQ are isogonal to each other wrt the angle CAB:

49

C A

Fig. 41 Proof of Theorem 19. a) (See Fig. 41.) Let O be the center of the circumcircle of triangle ABC: Then, O is the center of a circle through the points A; B; C; hence, by the central angle theorem, ]ABO = 90 ]BCA: The circumcenter O of triangle ABC obviously lies on the perpendicular bisector of its side BC: On the other hand, we know that the points P and Q lie on this perpendicular bisector. Hence, the points O; P; Q lie on one line, namely on the perpendicular bisector of the segment BC: Thus, OQ ? BC; so that ] (OQ; BC) = 90 : Hence, ]BQO = ] (BQ; OQ) = ] (BQ; BC) ] (OQ; BC) = ]QBC 90 ;

thus ]QBC = ]BQO + 90 : Since the point Q lies on the perpendicular bisector of the segment BC; we have BQ = CQ; this means that triangle BQC is isosceles, and satises ]QBC = ]BCQ: Hence, ]QBC = ]BQO + 90 becomes ]BCQ =

50

]BQO + 90 : Consequently, ]ACQ = ]BCQ ]BCA = (]BQO + 90 ) ]BCA = (90 ]BCA) + ]BQO = ]ABO + ]BQO = ]ABO ]OQB:

Consider the two lines OP and OB which intersect at the Euclidean point O; further, consider the two points P and Q on the line OP distinct from O; and the two points B and B on the line OB distinct from O: We can apply Theorem 18 to these lines and points. Assertion D1 of Theorem 18 then states that the triangles OP B and OBQ are oppositely similar, and Assertion D3 states that OP OQ = OB OB (where the segments are directed). As Theorem 18 ensures that the Assertions D1 and D3 are equivalent, we thus get: Auxiliary result 1. The triangles OP B and OBQ are oppositely similar if and only if OP OQ = OB OB: Now we can show the equivalence of the Assertions F1 and F5 : If Assertion F1 holds, then ]ABP = ]ACQ: Since ]ABP = ]ABO + ]OBP and ]ACQ = ]ABO ]OQB; this becomes ]ABO + ]OBP = ]ABO ]OQB; so that ]OBP = ]OQB: Furthermore, it is evident that ]BOP = ]QOB: Hence, the triangles OP B and OBQ are oppositely similar. According to Auxiliary result 1, this entails OP OQ = OB OB; that is, OP OQ = OB 2 : Now, we know that the points O; P; Q lie on one line, and that O is the center and OB is the radius of the circumcircle of triangle ABC: Hence, the equation OP OQ = OB 2 signies that the points P and Q are inverse to each other wrt the circumcircle of triangle ABC: Thus, Assertion F5 must hold. Conversely: Assume that Assertion F5 holds. This means that the points P and Q are inverse to each other wrt the circumcircle of triangle ABC: Since O is the center and OB is the radius of this circumcircle, this yields OP OQ = OB 2 : In other words, OP OQ = OB OB: According to Auxiliary result 1, we thus conclude that the triangles OP B and OBQ are oppositely similar, so that ]OBP = ]OQB: Hence, ]ABP = ]ABO + ]OBP = ]ABO ]OQB = ]ACQ; and thus Assertion F1 is valid. Thus we have shown that Assertions F1 and F5 are equivalent. In a similar way we can prove that Assertions F2 and F5 are equivalent. As seen above, ]QBC = ]BCQ: Thus, ] (BQ; CP ) = ] (BC; CP ) + ] (BQ; BC) = ]BCP + ]QBC = ]BCP + ]BCQ; so that ]ACP ]ABQ = ] (AC; CP ) ] (AB; BQ) = (] (AC; BQ) + ] (BQ; CP )) = ] (BQ; CP ) (] (AB; BQ) ] (AC; BQ)) = ] (BQ; CP ) = (]BCP + ]BCQ) ]BAC: ] (AB; BQ) ] (AB; AC)

Consequently, we have ]ACP = ]ABQ if and only if ]BCP + ]BCQ = ]BAC: In other words, Assertion F2 is equivalent to Assertion F3 : We have ]QBC = ]BCQ and similarly ]P BC = ]BCP: Thus, the equation ]BCP + ]BCQ = ]BAC is equivalent to the equation ]P BC + ]QBC = ]BAC: In other words, Assertion F3 is equivalent to Assertion F4 : 51

Altogether, we have proven the equivalence of all ve Assertions F1 ; F2 ; F3 ; F4 and F5 ; thus, Theorem 19 a) is veried. b) Assume that one of the ve Assertions F1 ; F2 ; F3 ; F4 and F5 holds. Since, according to Theorem 19 a), all these ve Assertions are equivalent, we can therefore conclude that Assertion F1 holds, i. e. we have ]ABP = ]ACQ: (See Fig. 42.) Since the point P lies on the perpendicular bisector of the segment BC; we have P B = P C: The circle with center P and radius P B = P C passes through the points B and C; let BP and CP be the points of intersection of this circle with the lines AB and CA dierent from B and C: Then, P BP = P CP = P B = P C: Since P BP = P B; the triangle BP BP is isosceles, so that ]BBP P = ]P BBP : Hence, ]ABP P = ]BBP P = ]P BBP = ]ABP = ]ACQ: On the other hand, the points B; C; BP and CP lie on one circle (namely, on the circle with center P and radius P B = P C); thus, ]BBP CP = ]BCCP : This yields ]ABP CP = ]BBP CP = ]BCCP = ]ACB: Similarly, ]ACP BP = ]ABC: Consequently, the triangles ACP BP and ABC are oppositely similar. From ]ABP P = ]ACQ and ]ABP CP = ]ACB; it follows that ]CP BP P = ]ABP P ]ABP CP = ( ]ACQ) ( ]ACB) = (]ACQ ]ACB) = ]BCQ:

Similarly, ]BP CP P = ]CBQ: Thus, the triangles P CP BP and QBC are oppositely similar. Since triangle ACP BP is oppositely similar to triangle ABC; and triangle P CP BP is oppositely similar to triangle QBC; the quadrilateral ACP P BP formed by the triangles ACP BP and P CP BP is oppositely similar to the quadrilateral ABQC formed by the triangles ABC and QBC: Consequently, ]CP AP = ]BAQ: In other words, ] (CA; AP ) = ] (AB; AQ) : Thus, the line AP is the isogonal of the line AQ wrt the angle CAB: This proves Theorem 19 b), and thus concludes the proof of Theorem 19.

52

B BP Q

C CP A
Fig. 42 Theorem 19 can be used to obtain a result about the isogonal conjugates of the so-called Kiepert points of the triangle. These points are dened as follows: (See Fig. 43.) Let ' be an arbitrary angle. On the sides BC; CA; AB of triangle ABC; we erect isosceles triangles BA' C; CB' A; AC' B with the bases BC; CA; AB and the equal base angle ]BCA' = ]A' BC = ]CAB' = ]B' CA = ]ABC' = ]C' AB = ': The Kiepert theorem states that the lines AA' ; BB' ; CC' concur at one point. We denote this point by K' ; and call it the '-Kiepert point of triangle ABC: Occasionally, triangle A' B' C' is referred to as the '-Kiepert triangle of triangle ABC:

53

K C

Fig. 43 0 0 (See Fig. 44.) Let A0' ; B' ; C' be the inverses of the points A' ; B' ; C' in the circumcircle of triangle ABC: Let O be the center of this circumcircle. Since triangle BA' C is isosceles with base BC; we have BA' = CA' ; thus, the point A' lies on the perpendicular bisector of the segment BC: The circumcenter O of triangle ABC also lies on the perpendicular bisector of its side BC: Thus, the line OA' is the perpendicular bisector of the segment BC: Now, the point A0' is the inverse of the point A' in the circumcircle of triangle ABC and therefore lies on the line OA' (since O is the center of this circumcircle). Thus, the point A0' lies on the perpendicular bisector of the segment BC: Now as we know that the points A' and A0' both lie on the perpendicular bisector of the segment BC; we can apply Theorem 19 to them. Since the points A' and A0' are inverse to each other wrt the circumcircle of triangle ABC; they fulll Assertion F5 of Theorem 19. Since, according to Theorem 19 a), the Assertions F3 and F5 are equivalent, they therefore also fulll Assertion F3 ; that is, we have ]BCA' +]BCA0' = ]BAC: Furthermore, according to Theorem 19 b), the validity of Assertion F5 implies that the lines AA' and AA0' are isogonal to each other wrt the angle CAB: Since ]BCA' = '; the equation ]BCA' + ]BCA0' = ]BAC becomes ' + 54

]BCA0' = ]BAC; thus ]BCA0' = ]BAC ': Since the point A0' lies on the perpendicular bisector of the segment BC; we have BA0' = CA0' ; thus, triangle BA0' C is isosceles with base BC; and this yields ]BCA0' = ]A0' BC: Combining this with ]BCA0' = ]BAC '; we obtain ]BCA0' = ]A0' BC = ]BAC ': Similarly, triangle 0 0 0 CB' A is isosceles with base CA and fullls ]CAB' = ]B' CA = ]CBA '; and 0 0 0 triangle AC' B is isosceles with base AB and fullls ]ABC' = ]C' AB = ]ACB ':

A B C K C' K' C A B'

A'

Fig. 44 Since the lines AA' and are isogonal to each other wrt the angle CAB; the line AA0' is the isogonal of the line AA' wrt the angle CAB: Now, the line AA' is the line AK' : Hence, the line AA0' is the isogonal of the line AK' wrt the angle CAB: 0 0 Similarly, the lines BB' and CC' are the isogonals of the lines BK' and CK' wrt the 0 0 angles ABC and BCA: Thus, altogether, the lines AA0' ; BB' ; CC' are the isogonals of the lines AK' ; BK' ; CK' wrt the angles CAB; ABC; BCA; and hence they concur at one point, namely at the isogonal conjugate of the point K' wrt triangle ABC: Summing up, we see: Theorem 20. Let ABC be a triangle, and ' an arbitary angle. On the sides BC; CA; AB of triangle ABC; we erect isosceles triangles BA' C; CB' A; AC' B with the AA0'

55

bases BC; CA; AB and the equal base angle ]BCA' = ]A' BC = ]CAB' = ]B' CA = ]ABC' = ]C' AB = ': Then, the lines AA' ; BB' ; CC' concur at one point: the '-Kiepert point K' of triangle ABC: 0 0 Let A0' ; B' ; C' be the inverses of the points A' ; B' ; C' in the circumcircle of 0 0 triangle ABC: Then, the triangles BA0' C; CB' A; AC' B are isosceles with the bases 0 BC; CA; AB and the base angles ]BCA0' = ]A0' BC = ]BAC '; ]CAB' = 0 0 0 0 0 0 ]B' CA = ]CBA '; ]ABC' = ]C' AB = ]ACB ': The lines AA' ; BB' ; CC' concur at one point, namely at the isogonal conjugate of the point K' wrt the triangle ABC: Using the theory of Tucker circles, we can show that this isogonal conjugate lies on the Brocard axis of triangle ABC; which is the line joining the circumcenter and the symmedian point of triangle ABC: References [1] Kiran S. Kedlaya, Geometry Unbound, version of 17 Jan 2006. http://www-math.mit.edu/~kedlaya/geometryunbound/ [2] Jan van Yzeren, Pairs of Points: Antigonal, Isogonal, and Inverse, Mathematics Magazine 5/1992, pp. 339-347. [3] R. A. Johnson, Advanced Euclidean Geometry, New York 1960. [4] R. A. Johnson, Directed Angles in Elementary Geometry, American Mathematical Monthly 24 (1917) Nr. 3, pp. 101-105. [5] Jos Carlos Chvez Sandoval (Hucht), MathLinks topic #64747: Iso...? [Reections in perpendicular bisectors]. http://www.mathlinks.ro/Forum/viewtopic.php?t=64747 [6] Jean-Pierre Ehrmann (JPE), Darij Grinberg et al., MathLinks topic #18423: A property of isogonal conjugates in a triangle. http://www.mathlinks.ro/Forum/viewtopic.php?t=18423 [7] Paul Yiu, Hyacinthos message #10533. http://groups.yahoo.com/group/Hyacinthos/message/10533?l=1 [8] Darij Grinberg et al., MathLinks topic #14092: Problem 5 (Second Day). http://www.mathlinks.ro/Forum/viewtopic.php?t=14092 [9] Victor Prasolov, Tochki Brokara i izogonaljnoe soprjazhenie [Brocard points and isogonal conjugation], Biblioteka "Matematicheskoe prosveshjenie" #4, Moscow 2000. http://www.mccme.ru/mmmf-lectures/books/books/books.php

56