The idea: three squares that add up to 1

An ellipsoid is the set of points (x, y, z) with

x2/a2 + y2/b2 + z2/c2 = 1,   a, b, c > 0.

All three terms are squares, so none is negative. They add up to 1, so none can be bigger than 1. That is the whole story: the surface is closed (no edge, no opening) and bounded (it fits in a box). It is the 3D version of the ellipse x2/a2 + y2/b2 = 1.

An ellipsoid with a = 2, b = 1.5, c = 1 drawn in 3D with the x, y and z axes. Purple dots mark where it meets the axes, at ±a on x, ±b on y and ±c on z, and the three semi-axes are drawn as thick segments from the centre labelled a, b and c
a, b and c are the distances from the centre to the surface along x, y and z: the semi-axes.

a, b, c are the semi-axes

Put y = z = 0 in the equation: x2/a2 = 1, so x = ±a. The surface meets the x-axis at (±a, 0, 0), the y-axis at (0, ±b, 0) and the z-axis at (0, 0, ±c). These six points are the intercepts (the purple dots on the canvas), and a, b, c are the semi-axes: half the length of the ellipsoid along each axis. Change one and the surface stretches along that axis only. (Demo 1: a, b, c are the semi-axes)

Another way to see it: the ellipsoid is the unit sphere u2 + v2 + w2 = 1 stretched by a along x, b along y and c along z (x = au, y = bv, z = cw).

Sphere, prolate and oblate spheroids

Which semi-axes are equal decides the shape:

semi-axesshapelooks like
a = b = c = rsphere x2 + y2 + z2 = r2a ball
a = b < cprolate spheroida rugby ball, an egg standing up
a = b > coblate spheroidthe Earth, an M&M
a, b, c all differenttriaxial ellipsoida pebble

With a = b every cut z = h is a circle, so the spheroid is round about the z-axis: it is what you get by spinning an ellipse about one of its axes. Spin it about the long axis and you get the prolate one, about the short axis the oblate one. (Demo 2: sphere → prolate → oblate spheroid)

Three surfaces side by side, all with a = b = 2: a sphere with c = 2, a tall prolate spheroid with c = 3 and a flat oblate spheroid with c = 1, each with its equation underneath
Keep a = b and change only c: c = a is a sphere, c > a a rugby ball, c < a a flattened Earth.

Cuts: ellipses that shrink to a point

Cut with the plane z = h: put h in for z and move it to the right side.

x2/a2 + y2/b2 = 1 − h2/c2

  • |h| < c: the right side is positive. Divide by it and the cut is an ellipse with semi-axes a√(1 − h2/c2) and b√(1 − h2/c2). It is biggest at h = 0 (semi-axes a and b) and shrinks as |h| grows.
  • |h| = c: the right side is 0, so x = y = 0. The cut is one point, the top (0, 0, c) or the bottom (0, 0, −c).
  • |h| > c: the right side is negative and two squares cannot add up to it. The plane misses the surface. (Demo 5: cut beyond c finds nothing)

Both semi-axes are multiplied by the same factor √(1 − h2/c2), so every cut z = h has the same shape (the ratio a : b) and only its size changes. For a = 2, b = 1.5, c = 1.5 at h = 1 the factor is √(1 − 1/2.25) ≈ 0.75 and the cut has semi-axes 1.49 and 1.12. (Demo 3: slide the cut z = h from −c to c)

Left: an ellipsoid with a = 2, b = 1.5, c = 1.5 and coloured cuts at z = 0, 0.75, 1.2 and the top point at z = 1.5. Right: the same cuts flat, seen from above, as nested ellipses that shrink to a point, each labelled with its height and semi-axes
Every cut z = h is the same ellipse scaled by √(1 − h²/c²): full size at h = 0, a point at h = c.

The other two axes work the same way. The cut x = h is the ellipse y2/b2 + z2/c2 = 1 − h2/a2 in the y-z plane, a point at x = ±a, nothing beyond. (Demo 4: cut x = h, from −a to a) The cut y = h gives x2/a2 + z2/c2 = 1 − h2/b2.

cutellipse whilesemi-axespoint at
x = h|h| < ab√(1 − h2/a2), c√(1 − h2/a2)x = ±a
y = h|h| < ba√(1 − h2/b2), c√(1 − h2/b2)y = ±b
z = h|h| < ca√(1 − h2/c2), b√(1 − h2/c2)z = ±c

Bounded, and its volume

Each square is at most 1, so |x| ≤ a, |y| ≤ b, |z| ≤ c: the ellipsoid sits inside the box 2a × 2b × 2c. It is the only quadric surface that is closed and bounded.

Stretching the unit ball by a, b and c multiplies every volume by abc, so the volume is V = (4/3)πabc; for a sphere, (4/3)πr3. The start state a = 2, b = 1.5, c = 1.5 has V = (4/3)π · 4.5 ≈ 18.85. There is no such simple formula for the surface area unless two semi-axes are equal.

Facts at a glance

propertyvalue
equationx2/a2 + y2/b2 + z2/c2 = 1
intercepts(±a, 0, 0), (0, ±b, 0), (0, 0, ±c)
centre, symmetrythe origin; symmetric in each of the three coordinate planes
cuts parallel to a coordinate planeellipse, a point, or nothing
boundedyes: |x| ≤ a, |y| ≤ b, |z| ≤ c
volume(4/3)πabc
general formx2/a2 + y2/b2 + z2/c2 − 1 = 0: three positive squares, constant −1

See also Quadric Surfaces for all nine shapes, Hyperboloid of One Sheet (flip the sign of z2) and Elliptic Cylinder (drop z2).

See also Cubic Surface, where raising two exponents of the sphere to 3 turns a bounded surface into one that runs off to infinity.

What the page leaves out

Tilted ellipsoids (cross terms xy, yz, zx; the overview page shows how a rotation removes them) and shifted ones ((x − x0)2/a2 + …). Cuts by tilted planes, which are also ellipses, and the two families of circular cuts of a triaxial ellipsoid. The surface area (elliptic integrals), the imaginary ellipsoid x2/a2 + y2/b2 + z2/c2 = −1 with no points, and uses such as the reference ellipsoid of the Earth (WGS 84) and the covariance ellipsoid in statistics.