You are on page 1of 17

Euclidean/non-Euclidean April 14: Spherical Geometry and

the Projective Plane: Hyperboloids


The surface of the unit sphere S in R3 is the graph of the equation
x 2 + y 2 + z 2 = 1. We get a di↵erential for arclength
p on S by restricting
the formula for arclength in R3 , namely ds = dx 2 + dy 2 + dz 2 (just as
we did for the unit circle in 2 dimensional space, but now with one more
variable). For any two points P andR Q on S, and any path from P to Q,
we get the length of the path as ds. The calculus of variations tells us
that the shortest path from P to Q, if when is the great circle path, that
is, the intersection of S with the plane through the origin that goes
through P and Q. This is the geodesic path, and just like in R2 , we take
this geodesic for the definition of the line containing P and Q in the sense
of incidence geometry. As we’ve seen, any two lines intersect, so spherical
geometry is elliptic. Sums of angles in triangles are greater than 180
degrees. STEM
p D Triangle has
1800
angle sum

April 14, 2021 1 / 17


Measuring distance on the sphere can be done by integration, using the
parameterization (spherical coordinates)

x = cos u, y = sin u cos v , z = sin u sin v .


However, just as in the case of a circle, it’s easiest to use the dot product.
Let P = (a1 , b1 , c1 ) and Q = (a2 , b2 , c2 ). Then the formula for distance is
1 1
dsphere (P, Q) = cos (P · Q) = cos (a1 b1 + a2 b2 + a3 b3 ),

which is the same formula as for the circle.

P o
or
spherical
stunt
d

April 14, 2021 2 / 17


Enter the Hyperboloid

The surface of the unit hyperboloid H in R3 is the graph of the equation

x2 y2 z 2 = 1.

It is a gotten by revolving the unit hyperbola x 2 y 2 = 1 in the xy -plane


about the x-axis. So it has 2 pieces, one with x 1 and one with x  1.
y

x O

April 14, 2021 3 / 17


Geodesics on H

We get a di↵erential for parclength on S by analogue of what we did for the


hyperbola, namely ds = |dx 2 dy 2 dz 2 |. As before, for any two
points P andR Q on H, and any path from P to Q, we get the length of
the path as ds. The calculus of variations tells us that the geodesic
path from P to Q is again the intersection of H with the plane through
the origin that goes through P and Q. This geodesic path (shall we call it
“a great hyperbola”?) is what we take for the definition of the line
containing P and Q in the sense of incidence geometry. Sums of angles in
triangles are less than 180 degrees.

angle
sum of a
triangle
out a
plane cuts is less 180
hyperbola
April 14, 2021 4 / 17
Measuring distance on the hyperboloid can be done by integration, using
the parameterization (hyperboloid coordinates)

x = cosh u, y = sinh u cos v , z = sinh u sin v .


However, just as in the case of the sphere, it’s easiest to use the dot
product. Let P = (a1 , b1 , c1 ) and Q = (a2 , b2 , c2 ). Then the formula for
distance is
dhyperboloid (P, Q) = cosh 1 (P ·H Q)
which is the same formula as the hyperbola, where by analogy we set the
hyperbolic dot product to be P ·H Q = a1 b1 a2 b2 a3 b3 .

April 14, 2021 5 / 17


Enter back the projective plane P2 (R)

Recall that the real projective plane P2 (R) is the set of equivalence classes
in R3 : {(a, b, c) ⇠ ( a, b, c)| 6= 0, (a, b, c) 6= (0, 0, 0)} which is to say,
lines through the origin. Every line through the origin in R3 hits the unit
sphere S in precisely 2 (antipodal) points, so we can think of P2 (R) as a
hemisphere. Even better, let’s recall how we wrote it as R2 [ P1 (R) as the
real plane plus a line “at infinity.” For every point on S other than on the
great circle x = 0, you can intersect a line through the point and the
origin with the tangent plane x = 1. Indeed, on P2 (R), if cos u 6= 0, the
point (cos u, sin u cos v , sin u sin v ) ⇠ (1, tan u cos v , tan u sin v ) so the last
2 coordinates give a point (tan u cos v , tan u sin v ) 2 R2 . In fact, setting
r = tan u, it is giving the point (r cos v , r sin v ) in polar coordinates, so we
get the whole real plane.

Oi'i
April 14, 2021 6 / 17
Lines in the projective plane P2 (R)

Recall that the lines in the projective plane P2 (R) are planes through the
origin. In other words, when we think of P2 (R) as the unit sphere S with
antipodal points dentified, the lines in the projective plane are precisely the
same as the lines in the sphere!. These planes are of the form
ax + by + cz for constants a, b, c 2 R. When we switch from the
projective plane to the affine plane by setting the x-coordinate to 1, we
get the corresponding line a = by + cz in R2 .

Ii

April 14, 2021 7 / 17


Let’s remember how switching from the projective plane to the affine plane
takes us from an elliptic geometry (with no parallel lines) to a Euclidean
plane, with unique parallels to a fixed line through a given point. The key
is that the whole projective plane does not map to the affine plane – we
have to leave out the line at infinity. So two lines in the projective plane
which intersect on the line at infinity turn out to be parallel when
considered in the Euclidean plane.
7 name
in E 7 name E
to

Ine at w
m

l in the affineplane 1122


So m and
their intersectsin
are parallel point ismissing
April 14, 2021 8 / 17
The projective plane P2 (R) versus the hyperboloid.

Points in P2 (R) are lines (a, b, c), 2 R, through the origin, and they
intersect the unit hyperboloid H precisely when a2 b 2 c 2 > 0, so the
hyperboloid is “one-half” (call it T ) of the projective plane. We can now
see how switching from the projective plane to hyperboloid takes us from
an elliptic geometry to a hyperbolic geometry, with multiple parallels to a
fixed line through a given point. The key is that one half the projective
plane, namely T , corresponds to points on H. Take any line ` on H and a
point P not on `. They correspond to a plane L through the origin and a
line P through the origin. Take any point Q 2 P2 (R) T (the other half),
which is a line Q through the origin. Let M be the plane through the
origin containing P and L. Then the plane M intersects H to gives a line
m on H, but M \ L is Q, which is not a point of H. So m is a parallel to `
on H through P. As we vary our choice for Q, we get infinitely many
parallels to ` through P.

April 14, 2021 9 / 17


Projecting from H to the unit disk D in R2

For every point on T you can intersect a line through the origin with the
tangent plane x = 1. Indeed, corresponding to
(cosh u, sinh u cos v , sinh u sin v , we get the point cao

(1, tanh u cos v , tanh u sin v ),

so the last 2 coordinates give a point (tanh u cos v , tanh u sin v ) 2 R2 . But
what kind of point? Note that
u

(tanh u cos v )2 + (tanh u sin v )2 = tanh u 2 < 1.

In other words, projecting T onto the tangent plane maps it one-to-one


(and smoothly) onto the interior of the unit disk D : y 2 + z 2 < 1. This is
the Beltrami-Klein model of H (which takes us from H to T to D).

April 14, 2021 10 / 17


We have the points of D: what are the lines?

Lines in the projective plane P2 (R) are planes through the origin, just as
they are in the hyperboloid. As before, planes are of the form ax + by + cz
for constants a, b, c 2 R. When we switch from T inside the projective
plane to D by setting the x-coordinate to 1, we get the line a = by + cz in
R2 . So lines in D are just ordinary lines in R2 restricted to D.

April 14, 2021 11 / 17


Distance between points on D
The formula for distance of two points P and Q on D is given by the
hyperbolic distance on H of the points that they come from.
Since the hyperboloid is a surface of rotation about the x-axis, we can
rotate those two points till they have the same y -coordinate: call it c.
Then the points lie on

x2 z 2 = 1 + c 2 = d 2,

say, which is just a scaled version of the unit hyperbola by the constant d,
and P and Q are of the form (1/d, tanh(u1 )/d), (1/d, tanh(u2 )/d) for
some u1 and u2 . We solved the problem of how to go from such points to
their distance d|u1 u2 | on Monday, which we called the projective
distance between 2 points on an interval (r , s).
The answer was
(s ↵)(r )
dproj (↵, ) = (1/2)| log | ||.
(r ↵)(s )

April 14, 2021 12 / 17


The way we interpret this is that if P, Q lie on a line `, which intersect the
unit circle at R and S. Then

(PS)(QR)
dproj (P, Q) = (1/2)| log | ||.
(PR)(QS)
This is the formula given for the Klein distance in Theorem 7.4 on page
343. S
o
Q
v
P
o

April 14, 2021 13 / 17


An example of calculating distances on D
Let’s apply the formula

(PS)(QR)
dproj (P, Q) = (1/2)| log | ||,
(PR)(QS)

when P = (1/2, 1/2), Q = (1/3, 2/3).


First we need the equation of the line ` between them, which by the
point-slope method is y = x + 1. This intersects the unit circle when

1 = x 2 +y 2 = x 2 +( x+1)2 = 2x 2 2x+1 =) 2x(x 1) = 0 =) x = 0, 1.

So R = (0, 1), S = (1, 0). By the Euclidean distance formula,


p p p p
PS = 2/2, QR = 2/3, PR = 2/2, QS = 2 2/3.

dproj (P, Q) = 1/2| log 1/2| = log 2/2.


.
April 14, 2021 14 / 17
Angles between rays on D

This is the chief problem with the Beltrami-Klein model: the Euclidean
angle lines make in D are not the same as the angle of the corresponding
rays on the hyperboloid H. In the Poincaré models these angles match up
(but lines are more complicated!)
However, one can compute the angle between two rays by making it a part
of a hyperbolic triangles and using the hyperbolic law of cosines!
Suppose A, B, C are three points in D, forming a triangle whose
(projective) lengths opposite vertices A, B, C are a, b and c. Then the
angle ✓ at vertex A satisfies:

cosh a = cosh b cosh c sinh b sinh c cos ✓.

B a
c
g C
A b

April 14, 2021 15 / 17


For Breakout

1) Verify the parameterization of H : x 2 y2 z 2 = 1 given above:

x = cosh u, y = sinh u cos v , z = sinh u sin v .


2) Compute the projective distance between the points (1/2, 1/2) and
(1/3, 0) on D.

April 14, 2021 16 / 17


For next class

Work on projects.

April 14, 2021 17 / 17

You might also like