Video 3, chapter 15

The Ellipse That Aims for You

Why a ball shot from one focus of an elliptical table always bounces through the other focus, explained by the reflection property of the ellipse, the foci of an ellipse, the law of reflection and elliptical billiards.

  • Video 3, chapter 15
  • Shapes in motion

Picture a billiard table shaped like an ellipse, with no pockets, and two special points inside it called the foci. Shoot a ball from one focus in any direction you like. After it bounces off the cushion, it travels straight through the other focus, whichever direction you chose. This page explains why by connecting the way an ellipse is drawn with a string to the rule that a bounce makes equal angles, and it states exactly which idealizations the claim needs.

Key terms

Reflection property of the ellipse

Definition Let an ellipse have foci F1F_1 and F2F_2, so that every point PP on it has PF1+PF2=2aPF_1 + PF_2 = 2a for one fixed length 2a2a, and fix such a point PP. The reflection property says that the segments PF1PF_1 and PF2PF_2 make equal angles with the tangent line at PP, the two angles lying on opposite sides of PP along the tangent. Equivalently, the line through PP perpendicular to the tangent splits the angle F1PF2F_1 P F_2 into two equal halves. In billiard terms, a ball that leaves F1F_1 in any direction and bounces off the ellipse by the law of reflection passes through F2F_2 after the bounce, on the ideal table described in the elliptical billiards card. The reason is the definition of the ellipse. Extend the segment from F1F_1 through PP to the point RR beyond PP with PR=PF2PR = PF_2, so that F1R=PF1+PF2=2aF_1 R = PF_1 + PF_2 = 2a. Let ℓ\ell be the perpendicular bisector of the segment F2RF_2 R, which passes through PP because PP is equally far from F2F_2 and RR. Every other point QQ of ℓ\ell is also equally far from F2F_2 and RR, and it does not lie on the straight line through F1F_1 and RR, which meets ℓ\ell only at PP. So

QF1+QF2=QF1+QR>F1R=2a.\begin{aligned} QF_1 + QF_2 &= QF_1 + QR \\[0.4em] &> F_1 R = 2a . \end{aligned}

Hence every point of ℓ\ell except PP lies outside the ellipse, and a line that touches the ellipse at PP and stays outside it everywhere else is the tangent at PP. Reflection across ℓ\ell swaps RR and F2F_2, so it carries the straight continuation PRPR of the incoming path onto the segment PF2PF_2. A mirror image makes the same angle with the mirror as the original, so the incoming path and PF2PF_2 make equal angles with the tangent, which is exactly what the law of reflection asks for.

In plain words At every point of an ellipse, the tangent is tilted just right to act as a mirror between the two foci. The tilt comes from the string. Sliding away from PP along the tangent always makes the two distances to the foci add up to more than the length of the string.

Example Take the ellipse

x225+y29=1\frac{x^2}{25} + \frac{y^2}{9} = 1

with foci F1=(−4,0)F_1 = (-4, 0) and F2=(4,0)F_2 = (4, 0), and the point P=(4,9/5)P = (4, 9/5) directly above F2F_2. The tangent at a point (x0,y0)(x_0, y_0) of this ellipse is x0x/25+y0y/9=1x_0 x/25 + y_0 y/9 = 1, which at PP becomes 4x+5y=254x + 5y = 25. To check that this line touches the ellipse only at PP, substitute y=(25−4x)/5y = (25 - 4x)/5 into 9x2+25y2=2259x^2 + 25y^2 = 225,

9x2+(25−4x)2=22525x2−200x+400=025(x−4)2=0,\begin{aligned} 9x^2 + (25 - 4x)^2 &= 225 \\[0.4em] 25x^2 - 200x + 400 &= 0 \\[0.4em] 25(x - 4)^2 &= 0 , \end{aligned}

which has the single solution x=4x = 4. The tangent falls with slope −4/5-4/5, so it makes an angle of about 38.66∘38.66^\circ with the horizontal. The segment PF2PF_2 is vertical, so it meets the tangent at 90∘−38.66∘=51.34∘90^\circ - 38.66^\circ = 51.34^\circ. The segment from F1F_1 to PP rises with slope 9/409/40, about 12.68∘12.68^\circ above the horizontal, and it meets the tangent at 12.68∘+38.66∘=51.34∘12.68^\circ + 38.66^\circ = 51.34^\circ. The angles agree, so a ball from F1F_1 that hits the cushion at PP goes straight down through F2F_2.

Foci of an ellipse

Definition Choose two points F1F_1 and F2F_2 in the plane, a distance 2c2c apart, and a length 2a2a with a>c≥0a > c \ge 0. The ellipse with foci F1F_1 and F2F_2 is the set of all points PP with

PF1+PF2=2a.PF_1 + PF_2 = 2a .

Points with a smaller sum lie inside the ellipse and points with a larger sum lie outside it. With the center at the origin and the foci at (±c,0)(\pm c, 0), the ellipse has the equation

x2a2+y2b2=1,b2=a2−c2.\begin{gathered} \frac{x^2}{a^2} + \frac{y^2}{b^2} = 1, \\[0.4em] b^2 = a^2 - c^2 . \end{gathered}

The long axis, the major axis, runs from (−a,0)(-a, 0) to (a,0)(a, 0) and contains both foci, and the short axis, the minor axis, runs from (0,−b)(0, -b) to (0,b)(0, b). When c=0c = 0 both foci sit at the center and the ellipse is a circle of radius aa. The condition a>ca > c matters, since a sum equal to 2c2c gives only the segment F1F2F_1 F_2 and a smaller sum gives no points at all.

In plain words Pin the two ends of a string of length 2a2a at the foci, pull it tight with a pencil and move the pencil around. The pencil draws the ellipse. As it moves, one piece of string gets longer and the other gets shorter by the same amount.

Example With a=5a = 5 and c=4c = 4 we get b=3b = 3, the ellipse x2/25+y2/9=1x^2/25 + y^2/9 = 1 with foci (±4,0)(\pm 4, 0). At (3,12/5)(3, 12/5) the distance to F1=(−4,0)F_1 = (-4, 0) is 49+144/25=37/5\sqrt{49 + 144/25} = 37/5 and the distance to F2=(4,0)F_2 = (4, 0) is 1+144/25=13/5\sqrt{1 + 144/25} = 13/5. Other points give the same total.

PointDistance to F1F_1Distance to F2F_2Sum
(0,3)(0, 3)55551010
(3,12/5)(3, 12/5)7.47.42.62.61010
(4,9/5)(4, 9/5)8.28.21.81.81010
(5,0)(5, 0)99111010

Law of reflection

Definition When an ideal ball, or a ray of light, hits a smooth wall at a point PP, it leaves along the path that makes the same angle with the tangent line of the wall at PP as the incoming path did, on the other side of the normal, the line through PP perpendicular to the tangent. Equivalently, the angle of incidence equals the angle of reflection, both measured from the normal, and the outgoing path is the mirror image, across the tangent line, of the straight continuation of the incoming path. The speed does not change. For a curved wall only the tangent line at the point of impact matters, so at that instant the wall acts like a flat mirror along its tangent. A path that meets the wall along the normal comes straight back, and running any reflected path backward gives another reflected path.

In plain words The wall turns the ball the way a mirror turns light, with the angle going in equal to the angle coming out. On a curved cushion, the mirror to use is the tangent line at the point of impact.

Example A ball travels from (0,3)(0, 3) to a straight wall along the xx-axis and hits it at (4,0)(4, 0). Without the wall it would continue to (8,−3)(8, -3). The mirror image of that point across the wall is (8,3)(8, 3), so the ball leaves toward (8,3)(8, 3), and both paths make an angle of about 36.87∘36.87^\circ with the wall, the angle whose tangent is 3/43/4. The route from (0,3)(0, 3) to (4,0)(4, 0) and on to (8,3)(8, 3) has length 5+5=105 + 5 = 10, the same as the straight segment from (0,3)(0, 3) to the mirror point (8,−3)(8, -3). Any other point of the wall gives a longer route, since its route has the same length as a bent path to (8,−3)(8, -3). The same mirror idea proves the reflection property of the ellipse.

Elliptical billiards

Definition Elliptical billiards is the mathematical model of a billiard table whose cushion is an exact ellipse. The ball is a point that moves in straight lines at constant speed, and at the cushion it reflects by the law of reflection with no loss of speed. There is no friction, spin, pocket, ball size or cushion that gives way. In this model the reflection property gives three facts.

  • A path through one focus passes through the other focus after the next bounce, then through the first focus again after the bounce after that, and so on.
  • Every path from F1F_1 to the cushion and on to F2F_2 has the same length 2a2a, so balls launched together from F1F_1 at the same speed reach F2F_2 together after one bounce.
  • A path that misses both foci on one stretch between bounces misses them on every stretch, because a single stretch through a focus would force all the earlier and later stretches through a focus as well.

Starting points other than the foci, and oval tables that are not ellipses, do not have this aiming property in general. On a real table the center of a ball of radius rr bounces off a curve lying rr inside the cushion, and that curve is not exactly an ellipse, so the exact argument no longer applies.

In plain words On a perfect elliptical table a shot from one focus is a sure shot through the other, however you aim it. The guarantee belongs to the two foci and to the exact shape of the cushion.

Example On the table x2/25+y2/9=1x^2/25 + y^2/9 = 1, a shot from F1=(−4,0)F_1 = (-4, 0) toward the top point (0,3)(0, 3) travels 55 to the cushion and then, by symmetry, 55 more down to F2=(4,0)F_2 = (4, 0). A shot toward (4,9/5)(4, 9/5) travels 8.28.2 to the cushion and then 1.81.8 straight down to F2F_2. A shot along the axis to the right passes F2F_2 after 88, hits (5,0)(5, 0) after 99, comes straight back and passes F2F_2 again after 1010. A shot to the left hits (−5,0)(-5, 0) after 11 and comes back to reach F2F_2 after 1+9=101 + 9 = 10. Every bounce path to F2F_2 has length 10=2a10 = 2a.

Check yourself

  1. A table has the shape x2/169+y2/25=1x^2/169 + y^2/25 = 1. Where are its foci? Find the length of the path from one focus off the cushion at (0,5)(0, 5) to the other focus, and the same for the cushion point (12,25/13)(12, 25/13).
Show the answer

Here a=13a = 13 and b=5b = 5, so c2=169−25=144c^2 = 169 - 25 = 144 and the foci are (±12,0)(\pm 12, 0). The point (0,5)(0, 5) is at distance 144+25=13\sqrt{144 + 25} = 13 from each focus, so the path has length 2626. The point (12,25/13)(12, 25/13) lies on the ellipse, since

144169+(25/13)225=144169+25169=1.\begin{aligned} &\frac{144}{169} + \frac{(25/13)^2}{25} \\[0.4em] &= \frac{144}{169} + \frac{25}{169} = 1 . \end{aligned}

It is directly above the focus (12,0)(12, 0), at distance 25/1325/13. Its distance to (−12,0)(-12, 0) is

242+(2513)2=31313,\sqrt{24^2 + \left(\frac{25}{13}\right)^2} = \frac{313}{13},

and 313/13+25/13=338/13=26313/13 + 25/13 = 338/13 = 26. Both paths have length 2a=262a = 26, as every one-bounce path between the foci does.

  1. Change a rule and start the ball at the center OO of the table x2/25+y2/9=1x^2/25 + y^2/9 = 1 instead of at a focus. In which directions does the ball pass through the center again right after its first bounce?
Show the answer

Only in the four directions along the axes. After a bounce at PP the ball can pass through OO again only by traveling back along the line POPO, which means it reverses. The law of reflection reverses a path only when the path meets the cushion along the normal, the line perpendicular to the tangent. The tangent at (x0,y0)(x_0, y_0) is x0x/25+y0y/9=1x_0 x/25 + y_0 y/9 = 1, so the normal points along (x0/25,y0/9)(x_0/25, y_0/9), and this is parallel to the direction (x0,y0)(x_0, y_0) of the segment from OO only when x0=0x_0 = 0 or y0=0y_0 = 0. So the shot must hit one of the four points (±5,0)(\pm 5, 0) and (0,±3)(0, \pm 3).

For example, a shot from OO toward (4,9/5)(4, 9/5) would continue to (8,18/5)(8, 18/5), whose mirror image across the tangent 4x+5y=254x + 5y = 25 is (128/41,−512/205)(128/41, -512/205). The ball leaves along the line toward that point, which crosses the major axis at x=3200/881≈3.63x = 3200/881 \approx 3.63, missing both the center and the focus (4,0)(4, 0). The proof of the reflection property needs the two distances from PP to add up to the same total all around the ellipse, and the foci are the only pair of points with that property.

  1. Extend the pattern. A ball leaves F1=(−4,0)F_1 = (-4, 0) on the table x2/25+y2/9=1x^2/25 + y^2/9 = 1 and first hits the cushion at (4,9/5)(4, 9/5). Find its next two bounce points and describe where the path goes after that.
Show the answer

After the first bounce the ball goes straight down through F2=(4,0)F_2 = (4, 0) and hits the cushion again at (4,−9/5)(4, -9/5). That stretch passed through F2F_2, so by the reflection property the next stretch passes through F1F_1. Its points are

x=4−8t,y=−95+95t,\begin{gathered} x = 4 - 8t, \\[0.4em] y = -\tfrac95 + \tfrac95 t, \end{gathered}

with t=0t = 0 at the bounce point and t=1t = 1 at F1F_1. Substituting into the equation of the ellipse and multiplying by 2525 gives 73t2−82t=073t^2 - 82t = 0. The root t=0t = 0 is the bounce just made, so the next bounce has t=82/73t = 82/73, at the point

(−36473,81365),\left( -\frac{364}{73}, \frac{81}{365} \right),

about (−4.986,0.222)(-4.986, 0.222). The path keeps alternating through the two foci, and the following bounces are at about (4.9998,−0.0247)(4.9998, -0.0247) and (−4.999998,0.0027)(-4.999998, 0.0027). From the third bounce on, the bounce points get closer and closer to the two ends of the major axis, and the path approaches the straight back and forth motion along that axis. This happens for every path through a focus that does not already run along the major axis.