This site is a work in progress. New lessons are added regularly. Contact us

Ellipses

Learning goals

  • Define an ellipse by a constant sum of focal distances
  • Square twice to reach x2a2+y2b2=1\tfrac{x^2}{a^2} + \tfrac{y^2}{b^2} = 1
  • Read b2=a2c2b^2 = a^2 - c^2 as a right triangle
  • Find the major axis under the larger denominator
  • Shift the center by replacing xx with xhx - h
  • Measure eccentricity e=cae = \tfrac{c}{a}, zero for a circle

Two pins and a string

The circle in the previous lesson was built from one fixed point: every point of the circle is the same distance rr from the center. An ellipse is built from two fixed points, called the foci (singular: focus). A point is on the ellipse when its distances to the two foci add to a constant.

The two-focus definition of an ellipseAn ellipse with foci F1 at (-c, 0) and F2 at (c, 0). Segments from a point P on the curve to each focus are labeled r1 and r2, and their sum is 2a.F1(-c, 0)F2(c, 0)P(x, y)r1r2r1 + r2 = 2a, the same for every P
Wherever the pencil P sits on the curve, the two pieces of string add to the same total. Slide P around and r1 and r2 each change, but r1 + r2 never does.

The distance condition

Set up coordinates so the two foci sit symmetrically on the xx-axis, at F1(c,0)F_1(-c, 0) and F2(c,0)F_2(c, 0) with c0c \ge 0. Call the constant total length of the string 2a2a. Writing the constant as 2a2a rather than as a single letter costs nothing now and saves a division by two later.

You need one tool to turn distances into algebra, and it is just the Pythagorean theorem. The segment from (x1,y1)(x_1, y_1) to (x2,y2)(x_2, y_2) is the hypotenuse of a right triangle whose legs are the horizontal gap x2x1|x_2 - x_1| and the vertical gap y2y1|y_2 - y_1|, so its length is (x2x1)2+(y2y1)2\sqrt{(x_2 - x_1)^2 + (y_2 - y_1)^2}. With that, the point P(x,y)P(x, y) lies on the ellipse exactly when

(x+c)2+y2r1  =  PF1  +  (xc)2+y2r2  =  PF2  =  2a.\underbrace{\sqrt{(x + c)^2 + y^2}}_{r_1 \;=\; PF_1} \;+\; \underbrace{\sqrt{(x - c)^2 + y^2}}_{r_2 \;=\; PF_2} \;=\; 2a.

Before doing any algebra, notice that aa and cc cannot be just anything. Any trip from F1F_1 out to PP and back to F2F_2 is at least as long as the direct trip from F1F_1 to F2F_2, which has length 2c2c. That is the triangle inequality, and it says r1+r22cr_1 + r_2 \ge 2c, with equality only when PP sits on the segment between the foci. So there are three cases:

  1. If 2a<2c2a < 2c, the string is shorter than the gap between the pins and no point satisfies the condition.
  2. If 2a=2c2a = 2c, the string is pulled perfectly straight and the “curve” collapses to the segment joining the foci.
  3. If 2a>2c2a > 2c, there is slack, and you get a genuine closed curve.

So a real ellipse forces a>c0a > c \ge 0. Keep that inequality in view: it is what makes the next step legal.

From the string to the equation

The distance condition is exactly x2a2+y2b2=1\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1#

Start from the definition and isolate one radical, because two square roots cannot be cleared at once:

(x+c)2+y2  =  2a(xc)2+y2.\sqrt{(x + c)^2 + y^2} \;=\; 2a - \sqrt{(x - c)^2 + y^2}.

This is exactly the routine from Solving Radical Equations: get one radical alone, square, and repeat if a radical survives. Squaring both sides and using (mn)2=m22mn+n2(m - n)^2 = m^2 - 2mn + n^2 with m=2am = 2a and n=(xc)2+y2n = \sqrt{(x - c)^2 + y^2}, whose square is (xc)2+y2(x - c)^2 + y^2, gives

(x+c)2+y2  =  4a24a(xc)2+y2+(xc)2+y2.(x + c)^2 + y^2 \;=\; 4a^2 - 4a\sqrt{(x - c)^2 + y^2} + (x - c)^2 + y^2.

Expand (x+c)2=x2+2cx+c2(x + c)^2 = x^2 + 2cx + c^2 on the left and (xc)2=x22cx+c2(x - c)^2 = x^2 - 2cx + c^2 on the right. The x2x^2, the c2c^2, and the y2y^2 appear on both sides and cancel, which is the whole reason for putting the foci symmetrically about the origin. What survives is

2cx  =  4a24a(xc)2+y22cx.2cx \;=\; 4a^2 - 4a\sqrt{(x - c)^2 + y^2} - 2cx.

Move the 2cx-2cx across, move the radical term across, and divide by 44:

a(xc)2+y2  =  a2cx.a\sqrt{(x - c)^2 + y^2} \;=\; a^2 - cx.

One radical is left and it is already alone, so square a second time. On the left, squaring aa and the radical gives a2[(xc)2+y2]a^2\left[(x - c)^2 + y^2\right]; on the right, (a2cx)2=a42a2cx+c2x2(a^2 - cx)^2 = a^4 - 2a^2cx + c^2x^2:

a2(x22cx+c2+y2)  =  a42a2cx+c2x2.a^2\left(x^2 - 2cx + c^2 + y^2\right) \;=\; a^4 - 2a^2cx + c^2x^2.

The term 2a2cx-2a^2cx appears on both sides and cancels. Gather the x2x^2 terms on the left and the constants on the right:

a2x2c2x2+a2y2  =  a4a2c2.a^2x^2 - c^2x^2 + a^2y^2 \;=\; a^4 - a^2c^2.

Factor a2c2a^2 - c^2 out of the x2x^2 terms and out of the right-hand side:

(a2c2)x2+a2y2  =  a2(a2c2).(a^2 - c^2)\,x^2 + a^2y^2 \;=\; a^2(a^2 - c^2).

Here is where a>ca > c earns its keep. Because a>c0a > c \ge 0, the number a2c2a^2 - c^2 is positive, so it is the square of a positive real number, and we may define b>0b > 0 by

b2=a2c2,equivalentlya2=b2+c2.b^2 = a^2 - c^2, \qquad\text{equivalently}\qquad a^2 = b^2 + c^2.

Substituting b2b^2 turns the equation into b2x2+a2y2=a2b2b^2x^2 + a^2y^2 = a^2b^2, and dividing by a2b2a^2b^2, which is not zero, leaves the standard form:

x2a2+y2b2=1.\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1.

One duty remains, and skipping it would be dishonest. Squaring is not reversible: from u=vu = v you may always conclude u2=v2u^2 = v^2, but from u2=v2u^2 = v^2 you may only conclude u=±vu = \pm v. That is precisely how radical equations manufacture extraneous solutions, and we squared twice. Each squaring is safe on the region where both sides are non-negative, and both sides are non-negative here. Even so, rather than track signs step by step, it is cleaner to check the converse outright.

Suppose (x,y)(x, y) satisfies x2a2+y2b2=1\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1. Then x2a21\frac{x^2}{a^2} \le 1, so xa|x| \le a, and y2=b2(1x2a2)y^2 = b^2\left(1 - \frac{x^2}{a^2}\right). Substitute that y2y^2 into the squared distance from PP to F2F_2, then use 1b2a2=a2b2a2=c2a21 - \frac{b^2}{a^2} = \frac{a^2 - b^2}{a^2} = \frac{c^2}{a^2} and b2+c2=a2b^2 + c^2 = a^2:

(xc)2+y2=x2(1b2a2)2cx+(c2+b2)=c2x2a22cx+a2=(acxa)2.\begin{aligned} (x - c)^2 + y^2 &= x^2\left(1 - \frac{b^2}{a^2}\right) - 2cx + \left(c^2 + b^2\right) \\ &= \frac{c^2x^2}{a^2} - 2cx + a^2 = \left(a - \frac{cx}{a}\right)^2. \end{aligned}

Since xa|x| \le a and 0c<a0 \le c < a, the quantity cxa\frac{cx}{a} lies between c-c and cc, so acxaa - \frac{cx}{a} is at least ac>0a - c > 0. It is positive, so taking the square root needs no absolute value: r2=acxar_2 = a - \frac{cx}{a}. Running the identical computation with cc replaced by c-c gives r1=a+cxar_1 = a + \frac{cx}{a}, also positive. Adding them,

r1+r2=(a+cxa)+(acxa)=2a.r_1 + r_2 = \left(a + \frac{cx}{a}\right) + \left(a - \frac{cx}{a}\right) = 2a.

Every point of x2a2+y2b2=1\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1 really does lie on the string. The two squarings invented nothing and lost nothing, so the equation and the distance condition describe the same set of points.

That computation pays a bonus. It showed that the distance from a point of the ellipse to the focus F2F_2 is r2=acxar_2 = a - \frac{cx}{a}, and xx ranges over [a,a][-a, a], so r2r_2 ranges over [ac,  a+c][\,a - c,\; a + c\,]. The closest a point of the ellipse ever gets to a focus is aca - c, and the farthest is a+ca + c. Both extremes happen at the vertices, the two ends of the long axis.

What aa, bb, and cc mean

Read the geometry straight off x2a2+y2b2=1\frac{x^2}{a^2} + \frac{y^2}{b^2} = 1.

Set y=0y = 0: then x2=a2x^2 = a^2, so x=±ax = \pm a. The points (±a,0)(\pm a, 0) are the vertices, and the segment joining them is the major axis, of length 2a2a. So the length of the string equals the length of the major axis, which is a fact worth remembering.

Set x=0x = 0: then y2=b2y^2 = b^2, so y=±by = \pm b. The points (0,±b)(0, \pm b) are the co-vertices, and the segment joining them is the minor axis, of length 2b2b.

Since b2=a2c2b^2 = a^2 - c^2 and c0c \ge 0, we get bab \le a: the ellipse really is at least as wide as it is tall when the foci are on the xx-axis. And a2=b2+c2a^2 = b^2 + c^2 is not a formula to memorize, it is a right triangle you can point at. Go from the center straight up to a co-vertex (0,b)(0, b), then across to a focus (c,0)(c, 0). The two legs are bb and cc, so the hypotenuse has length b2+c2=a2=a\sqrt{b^2 + c^2} = \sqrt{a^2} = a. The distance from a co-vertex to each focus is therefore exactly aa, and the two of them add to 2a2a, exactly as the string demands. The co-vertex is the one point where you can see the string length with no calculation at all.

The a, b, c right triangle inside an ellipseThe legs of the triangle are c, from the center to a focus, and b, from the center to a co-vertex. Its hypotenuse, from the focus to the co-vertex, has length a, the same as the semi-major axis.a² = b² + c²acbafocusfocusco-vertexvertex
The right triangle center, focus, co-vertex has legs c and b and hypotenuse a. That is the whole content of a squared equals b squared plus c squared, and it is why no side can beat the hypotenuse a.

Because aa is the hypotenuse of that triangle, no side can beat it: aba \ge b and a>ca > c, with a=ba = b precisely when c=0c = 0, which is the circle. So outside that one case aa is strictly the largest of the three, and that single observation kills most of the sign errors people make with ellipses.

Check your understanding

Where are the foci of x2169+y2144=1\frac{x^2}{169} + \frac{y^2}{144} = 1?

Answer choices

Which variable gets a2a^2

Nothing forced the foci onto the xx-axis. Put them at (0,c)(0, -c) and (0,c)(0, c) instead and run the same derivation with the roles of xx and yy interchanged. Every step is identical, so the result is identical with xx and yy swapped:

x2b2+y2a2=1,b2=a2c2,foci (0,±c).\frac{x^2}{b^2} + \frac{y^2}{a^2} = 1, \qquad b^2 = a^2 - c^2, \qquad \text{foci } (0, \pm c).

So there is one rule to internalize, and it is the source of more wrong answers in this topic than anything else:

a2a^2 is always the larger denominator, and the major axis runs along the axis of the variable that sits above it. The foci lie on that same axis, cc units from the center, where c2=(larger denominator)(smaller denominator)c^2 = (\text{larger denominator}) - (\text{smaller denominator}). (When the two denominators are equal there is no larger one: that is the circle, and it has no major axis.)

Notice you never have to remember which of aa and bb to subtract: cc is always found by subtracting the smaller denominator from the larger one. The reason is that cc is a leg of that right triangle and aa is the hypotenuse.

Worked example 1 Compare x225+y29=1\frac{x^2}{25} + \frac{y^2}{9} = 1 and x29+y225=1\frac{x^2}{9} + \frac{y^2}{25} = 1

These two equations use the same numbers, and they are different ellipses.

For x225+y29=1\frac{x^2}{25} + \frac{y^2}{9} = 1, the larger denominator 2525 sits under x2x^2, so a2=25a^2 = 25 and b2=9b^2 = 9, giving a=5a = 5 and b=3b = 3. The major axis is horizontal. Then

c2=a2b2=259=16c=4.c^2 = a^2 - b^2 = 25 - 9 = 16 \quad\Rightarrow\quad c = 4.

The vertices are (±5,0)(\pm 5, 0), the co-vertices are (0,±3)(0, \pm 3), the foci are (±4,0)(\pm 4, 0), and the eccentricity (defined below) is ca=45\frac{c}{a} = \frac{4}{5}. As a check, the distance from the co-vertex (0,3)(0, 3) to the focus (4,0)(4, 0) is 16+9=5=a\sqrt{16 + 9} = 5 = a, so the two focal distances from that point add to 10=2a10 = 2a.

For x29+y225=1\frac{x^2}{9} + \frac{y^2}{25} = 1, the larger denominator 2525 now sits under y2y^2, so a2=25a^2 = 25 still gives a=5a = 5, but the major axis is vertical. The value of cc is the same,

c2=259=16c=4,c^2 = 25 - 9 = 16 \quad\Rightarrow\quad c = 4,

but now everything is measured up and down: the vertices are (0,±5)(0, \pm 5), the co-vertices are (±3,0)(\pm 3, 0), and the foci are (0,±4)(0, \pm 4). The second ellipse is the first one turned on its side.

Check your understanding

Which axis is the major axis of x24+y249=1\frac{x^2}{4} + \frac{y^2}{49} = 1, and how long is it?

Answer choices

Moving the center off the origin

An ellipse does not have to be centered at the origin. From Transformations of Graphs you already know the move: replacing xx by xhx - h shifts the graph right by hh, and replacing yy by yky - k shifts it up by kk. Doing both to the standard form slides the whole ellipse so its center lands at (h,k)(h, k):

(xh)2a2+(yk)2b2=1(major axis horizontal),\frac{(x - h)^2}{a^2} + \frac{(y - k)^2}{b^2} = 1 \qquad\text{(major axis horizontal)}, (xh)2b2+(yk)2a2=1(major axis vertical).\frac{(x - h)^2}{b^2} + \frac{(y - k)^2}{a^2} = 1 \qquad\text{(major axis vertical)}.

Shifting cannot change any length, so aa, bb, and cc mean exactly what they meant before. Everything is now measured from the center: the vertices are aa units from (h,k)(h, k) along the major axis. The co-vertices are bb units from the center along the minor axis, and the foci are cc units from the center along the major axis.

From a general equation to standard form

A general equation like 9x2+4y2+54x8y+49=09x^2 + 4y^2 + 54x - 8y + 49 = 0 hides an ellipse. To expose it, complete the square in xx and in yy, exactly as in the Completing the Square lesson, and then divide so the right side becomes 11.

Worked example 2 Put 9x2+4y2+54x8y+49=09x^2 + 4y^2 + 54x - 8y + 49 = 0 in standard form

Group the xx terms and the yy terms, and move the constant to the right side:

9x2+54x+4y28y=49.9x^2 + 54x + 4y^2 - 8y = -49.

Factor the leading coefficient out of each group, since completing the square needs a bare x2x^2 and a bare y2y^2:

9(x2+6x)+4(y22y)=49.9\left(x^2 + 6x\right) + 4\left(y^2 - 2y\right) = -49.

Complete each square inside its parentheses. Half of 66 is 33, and 32=93^2 = 9; half of 2-2 is 1-1, and (1)2=1(-1)^2 = 1. Watch the bookkeeping: adding 99 inside the first parentheses really adds 99=819 \cdot 9 = 81 to the left side, and adding 11 inside the second really adds 41=44 \cdot 1 = 4. So add those same amounts to the right side:

9(x2+6x+9)+4(y22y+1)=49+81+4=36.9\left(x^2 + 6x + 9\right) + 4\left(y^2 - 2y + 1\right) = -49 + 81 + 4 = 36.

Write each group as a square:

9(x+3)2+4(y1)2=36.9(x + 3)^2 + 4(y - 1)^2 = 36.

Divide every term by 3636 so the right side is 11:

(x+3)24+(y1)29=1.\frac{(x + 3)^2}{4} + \frac{(y - 1)^2}{9} = 1.

Now read it off. The center is (3,1)(-3, 1), because x+3=x(3)x + 3 = x - (-3). The larger denominator, 99, sits under the yy term, so a2=9a^2 = 9, a=3a = 3, and the major axis is vertical; b2=4b^2 = 4, so b=2b = 2. Then

c2=a2b2=94=5c=5.c^2 = a^2 - b^2 = 9 - 4 = 5 \quad\Rightarrow\quad c = \sqrt{5}.

Measuring from the center: the vertices are (3,1±3)(-3, 1 \pm 3), that is (3,4)(-3, 4) and (3,2)(-3, -2); the co-vertices are (3±2,1)(-3 \pm 2, 1), that is (1,1)(-1, 1) and (5,1)(-5, 1); and the foci are (3,1±5)(-3, 1 \pm \sqrt{5}).

Eccentricity, and the circle hiding inside every ellipse

Two ellipses can both have a=5a = 5 and look nothing alike: one nearly round, one nearly flat. What distinguishes them is how far the foci sit from the center compared with the size of the curve. That ratio is the eccentricity:

e=ca.e = \frac{c}{a}.

Since 0c<a0 \le c < a, the eccentricity always satisfies 0e<10 \le e < 1. It also controls the shape completely, because c=eac = ea and therefore

b2=a2c2=a2e2a2=a2(1e2),ba=1e2.b^2 = a^2 - c^2 = a^2 - e^2a^2 = a^2\left(1 - e^2\right), \qquad \frac{b}{a} = \sqrt{1 - e^2}.

At the flat end, pushing ee toward 11 pushes cc toward aa, which squeezes bb toward 00 and flattens the ellipse against its major axis. At the round end something better happens. Take e=0e = 0, which means c=0c = 0. The two foci merge into a single point at the center, b2=a20=a2b^2 = a^2 - 0 = a^2 gives b=ab = a, and the equation becomes

x2a2+y2a2=1x2+y2=a2,\frac{x^2}{a^2} + \frac{y^2}{a^2} = 1 \quad\Longleftrightarrow\quad x^2 + y^2 = a^2,

a circle of radius aa. The definition degenerates just as gracefully: with F1=F2=OF_1 = F_2 = O, the condition r1+r2=2ar_1 + r_2 = 2a becomes 2PO=2a2 \cdot PO = 2a, that is PO=aPO = a, which is precisely “every point is distance aa from the center.” A circle is not a different kind of curve. It is the ellipse whose two foci have collided, and the circle lesson is the special case e=0e = 0 of this one.

Eccentricity controls the shapeThree ellipses of equal semi-major axis with eccentricities 0, 0.6 and 0.8. As the foci separate, the ellipse flattens.e = 0 (a circle)e = 0.6e = 0.8
Three ellipses with the same major axis. Only the foci move. At e = 0 the foci sit together at the center and the curve is a circle; as e grows the foci separate and the curve flattens.

Eccentricity is how astronomers describe orbits. Earth’s orbit has e0.017e \approx 0.017, so close to a circle that a drawing of it at page size is indistinguishable from one. Halley’s comet has e0.97e \approx 0.97, a long thin cigar of an orbit. Both are ellipses with the sun at a focus.

Check your understanding

What curve is x225+y225=1\frac{x^2}{25} + \frac{y^2}{25} = 1?

Answer choices

Building an ellipse from its geometry

Most problems hand you geometric facts (foci, a vertex, a string length, an eccentricity) and ask for the equation. The recipe never changes: find the center, decide which axis is major, then get any two of aa, bb, cc and use a2=b2+c2a^2 = b^2 + c^2 for the third.

Worked example 3 Foci (2,1)(2, -1) and (2,7)(2, 7), with focal distances summing to 1010

The center is halfway between the foci, so average the coordinates:

(2+22,  1+72)=(2,3).\left(\frac{2 + 2}{2}, \; \frac{-1 + 7}{2}\right) = (2, 3).

Both foci have the same xx-coordinate, so they lie on the vertical line x=2x = 2. The foci always lie on the major axis, so the major axis is vertical, which tells you a2a^2 belongs under the yy term.

The distance between the foci is 2c2c, and here that distance is 7(1)=87 - (-1) = 8, so c=4c = 4. The sum of the focal distances is 2a2a, and it is given as 1010, so a=5a = 5. Check the requirement a>ca > c: indeed 5>45 > 4, so a genuine ellipse exists. Now find bb:

b2=a2c2=2516=9.b^2 = a^2 - c^2 = 25 - 16 = 9.

Place the center at (2,3)(2, 3) with a2=25a^2 = 25 under the yy term:

(x2)29+(y3)225=1.\frac{(x - 2)^2}{9} + \frac{(y - 3)^2}{25} = 1.

Worked example 4 A semi-elliptical arch 3030 m wide and 1010 m high

A bridge arch is half an ellipse, cut along the major axis. Put the center of the full ellipse at the origin, with the road along the xx-axis. The arch spans 3030 meters, so the full major axis is 3030 and a=15a = 15. The arch is 1010 meters high at its middle, and that height is the semi-minor axis, so b=10b = 10. The upper half of

x2225+y2100=1\frac{x^2}{225} + \frac{y^2}{100} = 1

is the arch. How high is it 99 meters from the center? Substitute x=9x = 9:

81225+y2100=1y2100=1925=1625.\frac{81}{225} + \frac{y^2}{100} = 1 \quad\Rightarrow\quad \frac{y^2}{100} = 1 - \frac{9}{25} = \frac{16}{25}.

Multiply by 100100 and take the positive root, since the arch is the upper half:

y2=160025=64y=8.y^2 = \frac{1600}{25} = 64 \quad\Rightarrow\quad y = 8.

The arch is 88 meters high there. Note it is not 915\frac{9}{15} of the way down from 1010 to 00: the curve is not a straight line, and only the equation gets it right.

Common mistakes

Practice

Multiple Choice Questions (MCQ)

Progressively harder sets of questions. Each opens on its own page.

Free Response Questions (FRQ)

Longer questions in parts, to be worked out on paper. Progressive hints, the answer on its own so you can check yourself and try again, then the full worked solution, plus a rubric to mark your own work against.

Free response Work it out on paper 5 questions Start →
More practice (optional)

Extra sets, as hard as the Challenge set. Each one opens on its own page.

More resources (optional)

Other explanations of this lesson, if you want a second take.

A bit of history (Optional)

For two thousand years astronomers agreed on one thing. Whatever the heavens were doing, they were doing it with circles. A planet could ride a circle whose center rode a second circle, and a third could be stacked on if the fit went bad, but the parts were always circles. Mars was the planet that would not cooperate.

The measurements were the problem, because they were too good to explain away. Tycho Brahe, a Danish astronomer with the sharpest eyes of his century, had logged the position of Mars night after night for decades, all of it before the telescope. Johannes Kepler inherited those notebooks and spent years bending circles around them. His best attempt missed by eight minutes of arc, a quarter of the width of the full moon. Anyone else’s data could have absorbed an error that size. Brahe’s could not.

So Kepler threw the circles away. What fits, he announced in 1609, is an ellipse with the sun at one focus. He had named those two inner points himself, borrowing the Latin word for a hearth, because light or sound leaving one of them gathers at the other. That is the fireplace, and it is why a whisper crosses a domed room.

Now read the orbit with this lesson’s number. Mars has an eccentricity of about 0.090.09, so the curve is very nearly round, and drawn on a page it would pass for a circle. What betrayed it was the sun, sitting visibly off center. Push that number all the way down to zero and the two foci collide into one, which is the circle the astronomers wanted. Zero is the one value Mars would not take.