SCHOOL OF MATHEMATICS AND STATISTICS

MATH3511 Transformations, Groups and

Geometry
Term 3, 2019
Assignment 2 Solutions and Feedback

1. Since the three reflections each leave the incentre I invariant, T must also be a
reflection leaving I fixed, i.e. in a line through I.
But Ra sends the line AC to the line AB, Rb sends the line AB to the line CB
and Rc sends the line CB to the line CA, and thus T leaves the line AC invariant.
It follows that T is the reflection in the line orthogonal to AC through I. But the
intersection of this line and AC is the point where the incircle meets AC as radii are
orthogonal to tangents.
Alternatively for the last part, suppose the incircle touches the sides AC, AB and
BC at X, Y and Z respectively. Then (property of tangents) AX = AY , BY = BZ
and CZ = CX. So Ra sends X to Y , Rb sends Y to Z and Rc sends Z to X. Hence
T fixes X as well and so is a reflection in the line through I and X.
For an analytic proof, choose coordinates so that the origin is at the obvious fixed
point, I, vertex A lies on the negative x-axis and B is in the lower half plane. Thus
Ra is just reflection in the x-axis.

γ b
C

Ab b
I
α

β
b
B

Let 2α = ∠BAC, 2β = ∠ABC and 2γ = ∠BCA. The angle between the positive
x-axis and IB is then −(α + β) and that between the positive x-axis and IC is α + γ.
So the angle between IB and IC is 2α + β + γ = α + 21 π and Rc ◦ Rb is a rotation
about the origin through 2α + π.

1
 We can thus write the triple reflection as (see Q63 of problem sheet)
! !
cos(2α + π) − sin(2α + π) 1 0
Rc ◦ Rb ◦ Ra (x) = x
sin(2α + π) cos(2α + π) 0 −1
!
cos(2α + π) sin(2α + π)
= x
sin(2α + π) − cos(2α + π)

which is a reflection in a line at angle α + 12 π to the positive x-axis (Q63 again).

But this line is then through I and orthogonal to the side CA, and thus passes though
X, the tangency point of the incircle on that side, and the result follows.
COMMENTS Most of you got this out, one way or another: there were a couple of
other successful approaches too.
There were a fair number of you who seemed confused by the distinction between a
reflection and a rotation, and what was fixed by what. Of those who tried a coordi-
nate approach, picking the incentre as the origin was usually spotted, but then the
orientation of the axes was not used to simplify as much as was possible. This is an
important point I have mentioned several times: if you are free to choose the axes,
choose them to simplify as much as possible.
2. As the composition of two indirect isometries the composition is a direct isometry.
Choose coordinates so that ℓ is the x-axis and λ the y axis.
Then in these coordinates !
1 0
R(x) = x
0 −1
and !
−1 0
T (x) = x + αj
0 1
for some non-zero scalar α (and j the usual unit vector parallel to the y-axis).
Thus T ◦ R(x) = −x + αj and R ◦ T (x) = −x − αj. This a half turn (since it has
matrix part −I). From lectures and tutorials the fixed point is ± 21 αj respectively,
i.e. plus or minus (depending of the order of the composition) half the translation
component of T from the intersection of ℓ and λ.
Alternatively, for a more synthetic approach, since both R and T leave λ invariant,
the direct isometries R ◦ T and T ◦ R map λ to λ. This means they are either the
identity, translations in the direction of λ or half turns around a point on λ. The
compositions cannot be the identity, or in either case T would be the inverse of R,
and it clearly is not.
If either of them is a translation in the direction of λ, then they would also leave
invariant every line m parallel to λ. But every such line is orthogonal to ℓ and

2
 so invariant under R alone, i.e. R(m) = m. But if T ◦ R(m) = m then this says
T (m) = m, and if R ◦ T (m) = m, applying R to both sides also gives T (m) = m.
But T has only one invariant line, contradiction.
Thus R ◦ T and T ◦ R are half turns around a point on λ.
From lectures, a half turn around A is H(x) = −x+2a. Think of T as the composition
(in either order) of reflection in λ and a translation by (non-zero) u parallel to λ (u
is thus orthogonal to ℓ).
Let P be the point of intersection of ℓ and λ. Then
 
1
(T ◦ R)(P ) = T (p) = p + u = −p + 2 p + u
2
 
1
(R ◦ P )(P ) = R(p + u) = p − u = −p + 2 p − u
2

and comparing these to the standard formula tells us the fixed point is plus or minus
(depending of the order of the composition) half the translation component of T from
the intersection of ℓ and λ.
COMMENTS Most of you got this put, although there were a few who missed the point
of what a glide reflections is, and the descriptions of where the fixed points were was
sometimes not very clear.
And again, those who chose to use coordinates often did not choose them wisely.

3. We work in RP2 of course, so we do not have to worry about lines being parallel.
For each i, j with 1 ≤ i < j ≤ 3 let sides Ai Bi and Aj Bj meet at Pij , sides Bi Ci and
Bj Cj meet at Qij and sides Ci Ai and Cj Aj meet at Rij .
Desargues says that for any fixed ij, Pij , Qij and Rij are collinear (and define Lij ).
Consider the three triangles S1 = △P12 P13 P23 and S2 = △Q12 Q13 Q23 and S3 =
△R12 R13 R23 . Now if P12 , P13 and P23 are collinear (so that S1 does not exist) then
this line must contain all six vertices Ai and Bi . Similarly, if the Qij are collinear,
the line they are on contains all six vertices Bi and Ci . Thus if neither S1 nor S2
exist, all nine original vertices are collinear and our original triangles do not exist.
It follows that at least two of the triangles S1 , S2 and S3 must exist: let us assume
wlog it is S1 and S2 .
By construction, the sides P12 P13 and Q12 Q13 both pass through B1 . Similarly, the
sides P12 P23 and Q12 Q23 both pass through B2 and the sides P13 P23 and Q13 Q23 both
pass through B3 .
But B1 , B2 and B3 are collinear by our original assumption, and so S1 and S2 are in
perspective from this line. By Desargues this mean they are in perspective from a
point. But that says the lines P12 Q12 (i.e. L12 ), P13 Q13 (i.e. L13 ) and P23 Q23 (i.e. L23 )
are concurrent, as required.

3
 (Note that S1 and S3 (if they exist) will be in perspective from the line through the
Ak and S2 and S3 similarly in perspective from the line through the Ck .)
COMMENTS A lot of you got the key part of this out and many did it quite neatly.
Except, that is, that only one person spotted the issue of the triangles possibly not
existing (and they did not argue the case out successfully).
4. A projective transformation on RP2 can be written in homogeneous coordinates as
T (X) = AX for some invertible 3 × 3 matrix A.
   
X X
If T fixes the line at infinity then A  Y  = λ  Y  for some λ 6= 0 and all X, Y .
   

0 0
 
a 0 b
But this means A must have the form 0 a c  with ad 6= 0.
 

0 0 d
So in real coordinates the map is
a b a c
x′ = x + , y′ = y + .
d d d d
If a = dthis is a translation.
 Otherwise is a radial transformation with ratio a/d and
b c
centre , .
d−a d−a
COMMENTS Again, most people got this out.

5. Assuming the student number is 1234567, we want the lines through [2, 4, 6] and
[−1, 1, 1] and through [3, 5, 1] and [−2, 1, 0] Using Q84 from the problem sheets, the
homogeneous equation of the lines are
X Y Z X Y Z
2 4 6 = 0 and 3 5 1 =0
−1 1 1 −2 1 0
that is, respectively
−2X − 8Y + 6Z = 0 and − X − 2Y + 13Z = 0 ,
or (in real coordinates) x + 4y = 3 and x + 2y = 13.
The point of intersection has homogeneous coordinates [23Z, −5Z, Z] — it is thus
the real point (23, −5).
The ideal point on the first line is k[4, −1, 0] for k 6= 0.
COMMENTS A larger number of you than I would expect made a hash of calculating
the determinants here. Commonest loss of marks was just not giving me the ideal
point on the first line.