Reflection Around a Line In a Plane, Part 2

Now you understand the idea of reflections across a line – at least I assume that between the textbook, the teacher, and all the assigned homework you can work with reflections around the x-axis or the y-axis. But what if the line we want to reflect around isn't an axis? Or what if we don't have axes? Well, then we have a geometry problem.

Yes, geometry. Not algebra, not analysis.

This is way more fun.

Good browsers can display Scalable Vector Graphics.
Figure 0.

So here's the problem to be solved: Given an arbitrary figure, let's say a polygon, and an arbitrary line, find the reflection of that polygon around the arbitrary line. The problem is illustrated in Figure 0. ["Figure zero?" you ask. "Why zero?" Well, think about it. We're not to step 1; you haven't done anything yet.]

Specifically, we have a quadrilateral ABCD and a line r that doesn't intersect. And no coordinate system.

We could have a problem where r does intersect ABCD. Any point on line r would be its own reflection, and the rest of the solution would look just about the same.

Let's begin by reflecting the vertex A across the arbitrary line r. We have to start somewhere, so we might as well start with A.

We need a line through A that's perpendicular to r. (That's from definition of reflection.) We need to find the point A′ on that new line which is exactly the same distance from the line as A.

Fortunately, that's really easy to do. All we have to do is construct a line perpendicular to r through point A, which is something we've done before.

Draw a circle centered on point A and intersecting r in 2 points, J and K. This is shown in Figure 1. The circle can be any size you want, so long as the radius is bigger than the distance to the line r. (Otherwise you won't have J and K, and you need them.)

Good browsers can display Scalable Vector Graphics.
Figure 1.

If the circle were too small, it wouldn't reach to the line and there would be no intersection. If the radius were just exactly the distance to the line, the circle and line would intersect in only one point. (The line would be a tangent.)

We pick a bigger length for the radius so that we can be sure that it will intersect the line in 2 points.

Around each of the new points J and K, draw another circle. Make these circles the same size as the first. Two new circles, and all the same size.

The new circles will intersect in 2 points. Do you remember the theorem that says this? I hope so, because otherwise you'll have to stop now and prove it. We need those 2 points of intersection,

Notice the nice symmetry we created by using the same size circles! Both the new circles go through point A. (Of course they do. The distance from A to the point J is one radius; the distance back from J to A is also one radius. Same for K. What matters is that we used the same radius for all three circles.)

On the other side of line r, the circles also intersect. And it looks like a mirror image. Is it? Is that the point we want? Yes.

Good browsers can display Scalable Vector Graphics.
Figure 2.

If you have to prove that these 2 circles will intersect at 2 points, here are some hints for your proof.

Remember that J and K were both on the first circle, so the farthest apart they could be is one diameter (2 radii) – and that would only happen if they were in a line with A. Which they are not, since A is not on r.

There are 3 cases depending on the distance between centers.

Create triangles using the centers and the intersection points. You may need to use an indirect proof (suppose that there were a third point of intersection …).

It could be that J and K are closer than 1 radius, but the theorem is still valid.

The point we want is on a line that runs through A and is perpendicular to the line r (by the definition of reflection, remember). Also, the distance from A to the line r must be the same as the distance from the reflected point to r. Let's look at our points as vertices of triangles instead of intersections of circles. (It's just a different way to think about the same points we've been looking at.) Figure 2a shows this.

The line AA′ is perpendicular to r. (We know this from an earlier proof, but if you've forgotten it see the lemma.) So far, so good.

The distances to the line r must also be equal. If you previously proved the theorem that the bisector of the apex angle of an isosceles triangle also bisects the base of the isosceles triangle, that would establish that AP = PA′ Or, we can use our lemma again. It proves the congruence of triangles APJ and A′PJ (and APK and A′PK, too). That in turn implies the congruence of A′P and AP.

Good browsers can display Scalable Vector Graphics.
Figure 2a.

A Nice, Strong Lemma

Look at the triangles JAK and KA′J. Remember that JA, JA′, KA, and KA′ are all equal because they were created as radii of equal-sized circles. So these are both isosceles triangles. That means the base angles along line r are congruent: ∠KJA ≅ ∠JKA and ∠KJA′ ≅ ∠JKA′.

Besides that, AA′A′A by identity. So all the corresponding sides are congruent and the triangles are congruent to each other [SSS]. That means ∠JKA ≅ ∠JKA′ and ∠KJA′ ≅ ∠KJA.

Taking both results together, ∠KJA ≅ ∠JKA ≅ ∠KJA′ ≅ ∠JKA′.

Next, look at the other 2 triangles, KAA′ and JAA′. The same logic works for these; they are also isosceles and they are also congruent by SSS. So, ∠JA′A ≅ ∠JAA′ ≅ ∠KAA′ ≅ ∠KA′A.

Finally, look at the smaller triangles. For triangles AJP and AKP, JAKA because they started out as radii of equal-sized circles. We proved that ∠AJP ≅ ∠AKP and ∠PAK ≅ ∠PAJ. (OK, we used different names for the angles. But they are the same angles.) So AJP ≅ AKP by ASA.

For A′JP ≅ A′KP the argument is the identical. The argument is almost the same for AJP ≅ A′JP and AKP ≅ A′KP. So, all the little triangles are congruent to each other. This helps us in 2 ways:

  1. Because corresponding parts are congruent, ∠KPA ≅ ∠JPA ≅ ∠KPA′ ≅ ∠JPA′. If adjacent angles are equal, they are right angles (because they must add to 180°). That establishes that AA′ is perpendicular to line r.
  2. Because corresponding parts are congruent, AP = PA′. That establishes that A and A′ are the same distance from line r.

All that work, and so far we managed to reflect just the one point A. I can see how this might seem discouraging if it weren't so much fun.

Now that we know how to do it, we can easily reflect each of the vertices around line r. Figures 3 and 4 show how we find the reflection of point D.

First establish 2 points on r; I called them L and M this time. Then draw 2 circles centered at L and at M to find the reflection of point D.

Remember how using the same radius for all 3 circles let us find the line perpendicular to r (the line DD′) and the correct point on that line (D′) in a single step.

Good browsers can display Scalable Vector Graphics.
Figure 3.
Good browsers can display Scalable Vector Graphics.
Figure 4.

Repeat the process for points B and C.

Now if you connect the vertices you will have the entire reflected quadrilateral.

Good browsers can display Scalable Vector Graphics.
Figure 5.

Oh, oh! That last section contains a statement without any support. Did that bother you? Good! That's means you are thinking like a mathematician. You should never let your textbook or your teacher off the hook. Always demand careful reasoning for every claim.

The claim is that the non-vertex points on the sides of the quadrilateral ABCD will lie on the sides of the new quadrilateral A'B'C'D'. It seems obvious that this should be so, but how can we be sure?

Well, maybe you can let your teacher slide a little bit once in a while. But not too often.

Reflection Around a Line In a Plane, Part 2.5

Those of you who do not think like mathematicians may want to just skip this section. It isn't going to tell you anything that you don't think you already know; all we're doing is confirming that what we think is true really is true. If you don't care a whole lot about mathematical rigor, jump down to Part 3. It won't affect your grade, only your future.

But those of you who want to do math right should stick with us here. Part 2.5 will address those non-vertex points and show that the reflected quadrilateral really does look the way we think it ought to look.

Pick any side of the quadrilateral. This argument works with any of the sides, but pick AD, because that's what I picked (Figure 5a). Then make a trapezoid by connecting the side you picked with the corresponding side in the reflected quadrilateral.

AA′ and DD′ are segments of parallel lines because they are both perpendicular to line r. (Remember how we constructed them originally.) That fulfills the definition of a trapezoid.

Then pick a point along side AD. Whichever one you pick we'll call W.

Good browsers can display Scalable Vector Graphics.
Figure 5a.

If AD were parallel to line r we'd have a rectangle, which would make our proof easier, but for this particular quadrilateral we get all trapezoids.

Now add some more helper lines. Draw a segment from W that's perpendicular to r. (That will be parallel to the bases of the trapezoid, since all of them are perpendicular to r.) Where it meets the opposite side we'll call the point W′.

W′ might be the reflection of W since it is on the perpendicular line, but we don't know for sure yet. We also need to show equal distances from r, the line of reflection.

Next, drop altitudes from A and A′ to the opposite base. That will create a big rectangle AA′QR and some triangles. The altitudes have the same length because this is a trapezoid (or because AA′ || DD′). That is, AR = A′Q.

The rectangle has a constant width equal to AA′. (In particular, the opposite side QR has the same length, AA′.) Also, recall that line r bisects AA′ (because of how we constructed the figure). Then, too, because the sides are parallel to r, line r also bisects QR. That last point is interesting, because we constructed DD′ so that it, too, is bisected by r. By simple subtraction, you see that RD = QD′.

Good browsers can display Scalable Vector Graphics.
Figure 5b.

How do we know AA′QR is a rectangle? The bases of our trapezoid are parallel to each other and altitudes are pendendicular to the base, so all the angles are right angles.

In mathematics, you always need to be alert for when you prove something helpful. We just did. We've been emphasizing that all the helper lines are all perpendicular or parallel to each other. We also showed that AR = A′Q. and RD = QD′. Side-angle-side [SAS]! We just proved that the triangles ARD and A′QD′ are congruent.

But, you ask, how is that helpful? I know, nothing we do is ever obvious. This certainly isn't. All we really need is that ∠DAR ≅ ∠D′A′Q, which is true because those angles are corresponding parts of the congruent triangles.

Great. So now what? Now we can say that the little triangles AVW and A′UW′ are also congruent, using angle-side-angle [ASA]. We just showed that ∠DAR ≅ ∠D′A′Q. You have a right angle at U and at V (because the lines are perpendicular). The sides between those angles (AV and A′U) have equal lengths because WW′ is parallel to the sides of the rectangle.

But, you ask yet again, how is that helpful? That's helpful because UW′ = VW. (We know this now because they are corresponding sides of congruent triangles.)

Yes, yes, but how is that helpful? I hear some impatience in your voice. But look: This means that W and W′ are the same distance from line r. Just reverse the discussion about D and D′. UV is bisected by line r and now we know that the extra lengths VW and UW′ are equal. So, by simple addition, the distance from r to W is equal to the distance from r to W′ And that is what we still needed to demonstrate.

The reflection of any point W on the segment AD is the corresponding point on the constructed segment A′D′.

Now repeat this argument for each side of ABCD and you're done.

Aren't you glad that you insisted on a rigorous presentation?

Reflection Around a Line In a Plane, Part 3

Pure geometry is fun, but the real power of the mathematics is seen when geometrical thinking is combined with algebraic skills. In other words, when we put the geometry onto the coordinate plane.

Good browsers can display Scalable Vector Graphics.
Figure 6.

So let's look at our original problem with coordinate axes present. In Figure 6, there are gridlines in the background for reference (at 10, 20, 30, ...). More importantly, we have coordinates for each point A, B, C, and D and we have an equation for the line r. Otherwise, this is the same as Figure 0.

The problem, as before, is to find the reflection of ABCD around the line r.

The problem shown in Figure 6 is the same as the problem shown in Figure 0. The difference lies in how we look at the problem.

Let's start by looking at a point and thinking about what we know, geometrically, about our problem. We'll start with point A again, but our logic would apply to any point and its reflection around a line.

We know there will be a point A′ which is the reflection of A. We need to know the coordinates of point A′. (We know approximately where to draw this reflection based on our construction in Part 2.)

Because A′ is a reflection of A, we know that line r is a perpendicular bisector of the segment AA′. If we name the intersection point P, as before, we can say AP = PA′.

Next we construct 2 right triangles as shown in Figure 6a. Starting from A we move horizontally until we are at point T directly under P. Then move vertically up to P. From A′ we move to T′ directly above P, then down to P.

∠PAT ≅ ∠PA′T′ because they are formed by the intersection of the line AA′ with 2 parallel lines. ∠APT ≅ ∠A′PT′ because they are opposite angles of the intersection of AA′ and TT′. We already noted that AP = PA′. This gives us a 2 pairs of angles and a corresponding side between them. So, triangles ATP and A′T′P are congruent by angle-side-angle [ASA].

Good browsers can display Scalable Vector Graphics.
Figure 6a.
You can make the same argument analytically. This is really just the midpoint problem. I didn't want to use the midpoint formula because I like geometry more than algebra. (Even though algebra is fun, too.)
We know these are right triangles because the legs are parallel to the x and y axes.

Corresponding parts of congruent triangles are congruent. In particular, AT = A′T′ and PT = PT′. That means that however far you move horizontally from A to get to P, you have to move the same amount again to get from P to A′ – and however far you move vertically from A to get to P, you have to move the same amount again to get from P to A′.

That could be useful if we knew the coordinates of P. Can we know them? Yes.

Our line r has the equation
y = 2x
The equation for the line AA′ is
y = (-1/2)(x - 30) - 20.
Solving the 2 equations for x and y gives the coordinates of the point P.

Substituting into the second equation:
2x = (-1/2)(x - 30) - 20
4x = -1(x - 30) + 40
5x = 70
x = 14

Substituting back into the first equation:
y = 28

The coordinates of P are (14, 28).

Going from A to P, we moved horizontally from 30 to 14, a distance of -16. Vertically, we moved from 20 to 28, a distance of 8. So to get from P to A′ we move the same amounts again: 14 - 16 = -2 and 28 + 8 = 36.

The coordinates of A′ are (-2, 36).

In general, the equation of the line of reflection is
y = mx + b.

The equation for a perpendicular line through a known point (x0, y0) is
y = (-1/m)(x - x0) + y0.

Solving the 2 equations for x and y gives the coordinates of the intersection (point P).

Knowing the coordinates for both the original point and the intersection with line r, you can compute the coordinates of the reflected point.

The power of analytic geometry is that once we have determined the process we can obtain all the coordinates just by calculating. Here I summarize the calculations for all 4 points:

Original Intersection Reflection
A (30,20) (14,28) A′ (-2, 36)
B (70,10) (18,36) B′ (-34,62)
C (40,-60) (-16,-32) C′ (-72,-2)
D (20,-30) (-8,-16) D′ (-36,-2)

Knowing the coordinates, we can plot the reflected quadrilateral directly into the coordinate plane. This is shown in Figure 7.

Good browsers can display Scalable Vector Graphics.
Figure 7.

Valid HTML

Pivot Rock Ensign