Rose-Hulman Undergraduate Mathematics Journal

Volume 3
Article 3
Issue 1

Geodesics Using Mathematica

Jacob Lewis
Columbia University, jml104@columbia.edu

Follow this and additional works at: https://scholar.rose-hulman.edu/rhumj

Recommended Citation
Lewis, Jacob (2002) "Geodesics Using Mathematica," Rose-Hulman Undergraduate Mathematics Journal: Vol. 3 : Iss. 1 , Article 3.
Available at: https://scholar.rose-hulman.edu/rhumj/vol3/iss1/3
 1

Geodesics Using Mathematica

Jacob Lewis*

Columbia University

Abstract: We describe surfaces and geodesics without assuming prior

knowledge of differential geometry. This involves selecting and presenting basic

definitions and theorems. Included in this discussion are definitions of surface,

coordinate patch, geodesic, etc. This summary closes with a proof of the length-

minimizing properties of geodesics. Examples of surfaces are given and plotted

in Mathematica. We also describe geodesics on these surfaces and plot select

examples. The surfaces chosen include some with Clairaut patches, some

without, and some surfaces in R3 and some not in R3.

*
The author wishes to thank Professor Charles Doran for his gracious help.
 2

A well-known adage is, “The shortest distance between two points is a

straight line.” This is certainly true on the plane, but on other surfaces the adage

proves to be false. For example, assume the earth to be a sphere. New York

City and Madrid, Spain are both at latitudes of about 40°N. Yet an airplane

taking the shortest distance between the two does not follow the 40th parallel.

Rather, it arcs north, following the great circle (i.e., circle centered at the sphere’s

center) between the two cities.

To generalize the adage--and along the way to explain why planes travel

this way--we will introduce a special class of curves on surfaces, called

geodesics. Geodesics have the useful property that the shortest curve segment

connecting two points on a surface is a segment of a geodesic. As we shall see,

great circles are geodesics on the sphere, and they therefore have the property

that they are the "shortest" curves on the sphere. To examine geodesics, we will

develop connections between differential geometry, differential equations, and

vector calculus. In order to see geodesics, even when they cannot be found

explicitly, the computer algebra system Mathematica will be used.

Differential Geometry

A surface in three-dimensional Euclidean space (R³) is a set of points in

R³ that locally look like a plane-that is, given any point on the surface, there is a

small neighborhood of that point which appears to be planar. Again, the earth’s

surface taken as a sphere is a good example. The earth’s surface curves, yet by

looking around, one cannot see this curvature. This is because the area of the

earth one can see is a small enough neighborhood of the point where he/she is
 3

standing that this neighborhood appears flat. So the sphere is a surface in R³.

More technically,

Definition: M ⊂ R³ is a surface if for any x ∈ M, there exists an open

neighborhood U ⊂ R³ containing x, an open neighborhood W ⊂ R², and a
map x: W → U ∩ M that is differentiable with differentiable inverse. Such
a function is called a parameterization or a coordinate patch since it allows
us to assign coordinates to the surface corresponding to the coordinates
of R².

Surfaces do not need to be in R3; replacing R3 with a general space X yields a

valid definition. In fact, some of the surfaces we will examine cannot be placed in

R3.

For a sphere of radius r centered at the origin, a coordinate patch is x(u, v)

= (rcos(u)cos(v), rsin(u)cos(v), rsin(v)), for –π < u < π, and –π < v < π. A

coordinate patch is said to be orthogonal if its first partial derivatives are

orthogonal-that is, if xu•xv = 0. Clearly, for an orthogonal patch x, xu x xv is never

xu × x v
zero. This means it is possible to construct a unit normal U = at any
xu × x v

point on the surface. Also, because xu and xv vary smoothly on M, so will U. If

any two points on a surface can be connected by a curve contained in the

surface, the surface is said to be connected.

Using the sphere as an example, the parametrization of the sphere given

above is orthogonal. xu = (-r sin(u) cos(v), r cos(u) cos(v), 0), and xv = (-r cos(u)

sin(v), -r sin(u) sin(v), r cos(v)). So xu • xv = r2 sin(u) sin(v) cos(u) cos(v) - r2

sin(u) sin(v) cos(u) cos(v) + 0 = 0. The unit normal at x(u, v) is

 4

1
(cos 2 (u ), cos( u ) sin( u ), sin( v ))
cos (u ) + sin ( v )
2 2
, as the reader can easily verify

either by hand or using Mathematica.

A useful and important construct on a surface M is the tangent plane to

the surface at a point p. Parameterize M in a neighborhood of p by x(u, v), with

x(u0, v0) = p. Then, the tangent plane to M at p-denoted by TpM-is the two

dimensional vector space spanned by {xu(u0, v0), xv(u0, v0)}. It is fairly easy to

show that this space is equivalent to the space of all vectors v such that v =

α’(t0), where α is a curve on M with α(t0)=p (see [3]). Since TpM is a vector space,

an inner product can be defined on it. If an inner product is defined consistently

on every tangent plane of M, then M is said to be a geometric surface.

The Geodesic Equations

Important curves on surfaces are curves called geodesics. Geodesics are

essentially the extensions into M of straight lines in the plane—that is, relative to

the surface, there appears to be no acceleration. Formally,

Definition: For a surface M in Euclidean three-space, a geodesic is a

curve α:[0, 1]→M where α’’ is always normal to M.

Since α’’ is always normal to M, that means that the dot product of α’’ and a

vector in in TpM is always zero. In particular, α’’ • α’ = 0. Let s(t) be the speed of

α at t; s(t) = α ′(t ) • α ′(t ) . Differentiating s with respect to t,

α ′′(t ) • α ′(t )
(1) s′(t) = =0
α ′(t ) • α ′(t )

This shows that s(t) is a constant function, and so a geodesic must be a

constant-speed curve. In fact, we can parameterize any geodesic so that it is

 5

unit-speed.

Given an orthogonal coordinate patch x in a geometric surface M,

geodesics can be defined by differential equations called, appropriately, the

geodesic equations. Consider a curve α in M. Express α(t) = x(u(t), v(t)). Then,

∂x ∂x
α′= u ′( t ) + v ′( t ) = x u u ′ + x v v ′
∂u ∂v , and so

(2) α ′′ = xu u′′ + u′(xuu u′ + x uv v′) + x v v′′ + v′(x uv u ′ + x vv v′)

If α is a geodesic, then it is normal to the surface, and hence

(3) α″•xu = 0 and α″•xv = 0

So, using (2) and (3) and the fact that xu•xv = 0 yields the differential system

(4) |xu|2 u′′ + u′2 xuu•xu + 2u′v′xuv•xu + v′2xvv•xu = 0

(5) |xv|2 v′′ + v′2 xvv•xv + 2u′v′xuv•xv + u′2xuu•xv = 0

which a curve must satisfy to be a geodesic. These two equations are called the

geodesic equations, because solving them gives geodesics. In fact, these

equations allow us to generalize the definition of a geodesic to a surface that is

not in R3. For such a surface, we merely define a geodesic to be a solution of

the geodesic equations. An immediate result of this system of differential

equations is the following theorem:

Theorem: Given a surface M in R3 with a unit normal, a point p∈M, and

vector v∈TpM, there exists a unique geodesic γ such that γ(0)=p and
γ′(0)=v.

Proof: Let γ(t)=x(u(t), v(t)). Then γ(0)=p gives initial conditions u(0) and

v(0). γ′(0)=v gives initial conditions u′(0) and v′(0). Then, by the
 6

fundamental existence and uniqueness theorems of ordinary differential

equations [1], γ exists and is unique. !

If every geodesic can be extended infinitely without leaving the surface, then the

surface is called a complete surface.

Usually, the geodesic equations cannot be solved by hand. For this

reason, it is useful to be able to solve the geodesic equations numerically; we

give a Mathematica procedure for this below. On a few surfaces, such as the

sphere, explicit solutions can be found. One class of surfaces for which the

geodesic equations can often be explicitly solved are surfaces given a Clairaut

patch. A Clairaut patch is an orthogonal patch in which |xu| and |xv| are both

∂
independent of v. This implies (x u • xu ) = 2x uv • x u = 0 and
∂v

∂
(x v • x v ) = 2x vv • x v = 0 , or simply xuv•xu = xvv•xv = 0. Also, since xu•xv = 0,
∂v

∂
(x u • x v ) = x uu • x v + x u • x uv = x uu • x v = 0 . Therefore, the geodesic equations
∂u

reduce in this case to

(6) |xu|2 u′′ + u′2 xuu•xu + v′2xvv•xu = 0

(7) |xv|2 v′′ + 2u′v′xuv•xv = 0

Isolating the v-terms in (7) gives

d 2v  ∂G 
 
(8) dt = v ′′ = −2u ′ x uv • x v = − du  ∂u  , where G= x •x .
2
v v
dv v′ xv • xv dt  G 
 
dt  
 7

Integrating both sides of (8) with respect to t, and making the substitution w =

dw G dG
dv/dt, ∫ w
= − ∫ u du = − ∫
G G
⇒ ln v′ = − ln G + c . Taking the exponential of both

sides gives the equation

dv k
(9) = v′ =
dt G

for some constant k. Now we use the fact that we may assume the geodesic α(t)

= x(u(t), v(t)) to be unit speed. This means that

k2
1 = α ′ • α ′ = (x u u ′ + x v v′) • (x u u ′ + x v v′) = xu u ′ + Gv′ = xu u ′ +
2 2 2 2 2
, or
G

du G − k2
(10) = u′ = ± 2
dt xu G

2
dv xu k 2
Dividing (9) by (10) gives =± , and integrating this yields
du G2 − k 2 G

2
xu k 2
(11) v = ±∫ du
G2 − k 2 G

If for a particular surface with a Clairaut patch, the integral on the right-hand side

of (11) can be solved explicitly, this gives an explicit solution to the geodesic

equations for this surface. If |xu| and |xv| are independent of u, then the patch is

said to be v-Clairaut, and equation (11) holds by interchanging u and v. In this

case, we use the variable E instead of G.

Length Minimizing Properties of Geodesics

An important property of geodesics, alluded to earlier, is that if a shortest

curve between two points on a surface exists, it is a segment of a geodesic. (A

 8

shortest curve between two points might not exist if the surface is not

geodesically complete.) To show this, we make the following definition:

Definition: A curve segment α is a shortest segment from p to q if for any

other curve segment β from p to q, the length L(β)≥L(α). α is the shortest

segment from p to q if for any other curve segment β from p to q,

L(β)=L(α) implies that β is a reparameterization of α.

Now, for any point p on a surface M, there exists a small neighborhood N(p) of M

around p for which we can give a patch x(u,v), called the geodesic polar patch,

with the following properties: x is orthogonal, E = xu•xu = 1, and G = xv•xv > 0.

Details of how to define this patch are given in [4]. For q in N(p), there is a

geodesic segment connecting q and N(p) which is contained completely in N(p).

If q = x(u0, v0), then this geodesic, called the “radial geodesic,” is given by g(u) =

x(u, v0). Now,

Theorem: For each point q in N(p), the radial geodesic from p to q is the

shortest curve from p to q.

Proof: Let x be the above-mentioned parameterization of N(p). Let q=(u0,

v0). Then, γ(u) = x(u, v0). Let α be an arbitrary curve from p to q. Assume

L(α)≤L(γ). Without loss of generality, α contains no loops, for if it did we could

discard the loops, thereby shortening α. Therefore, we can write α(u) = x(a1(u),

a2(u)), with 0 ≤ u ≤ u0. Since x(a1(0), a2(0)) = x(0, 0) = p, and x(a1(u0), a2(u0)) =

x(u0, v0) = q, we see that a1(0)=a2(0)=0, a1(u0)=u0, and a1(v0)= v0. Since E=1

and the patch is orthogonal,

(5) α = Ea1′ 2 + Ga 2′ 2 = a1′ 2 + Ga 2′ 2 ≥ a1′ 2 = a1′ , so

 9

u0 u0

L(α ) = ∫ (a1′ ) + G (a 2′ ) du ≥ ∫ a1′du = u 0

2 2
(6)
0 0

u0

Now γ has unit speed, so L(γ ) = ∫ du = u 0 , so L(α)≥L(γ). If L(α)=L(γ), then

0

u0 u0

∫ a1′ + Ga 2′ du = ∫ a1′du = u 0 , so
2 2
(7)
0 0

a1′ 2 + Ga ′22 = a1′ . Then, since G > 0, a2’ = 0. So a2 = v0, which implies that α(u) =

x(a1(u), v0), which is a reparametrization of γ. So γ is by definition the curve of

smallest length. !

This remarkable theorem gives an idea of why visualizing geodesics is so

important to understanding the geometry of a surface. Just as lines on the plane

give the shortest path between two points on the plane, geodesics perform this

function on arbitrary surfaces. In a very real sense then, geodesics are the

abstraction of the notion of a line to other surfaces. (Indeed, the theorem implies

that straight lines are the geodesics on the plane.)

Mathematica and Geodesics

To be able to see geodesics on a surface—particularly useful in the

absence of an explicit equation—we turn to the computer algebra system

Mathematica. Mathematica can solve differential equations numerically using a

built-in function named NDSolve, and it can graph a parametrized surface using

a function named ParametricPlot3D. Note the capitalization; Mathematica

commands are case-sensitive. The definitive work on using Mathematica to

study surfaces is [2], which includes a handy appendix giving the Mathematica
 10

code necessary to make many surfaces and curves, and to do various other

things. The discussion here is intended to streamline and generalize the

methods described in [2] for dealing with geodesics.

To graph a surface with patch x(u, v) in Mathematica is a relatively simple

matter. First, input the equation for the patch. To input a function, the argument

goes in square brackets, an underscore follows the variables’ names, and a

colon precedes the equals sign. For example, to define the sphere, the

command is: sphere[u_,v_]:={r Cos[u] Cos[v], r Sin[u] Cos[v],

r Sin[v]}. Second, select an appropriate rectangle P = [u0, u1] x [v0, v1] in R2

on which to graph. Lastly, graph using the built-in function ParametricPlot3D.

The syntax is as follows: ParametricPlot3D[x[u, v], {u, u0, u1},

{v, v0, v1}, Lighting->True]. (The lighting is important; otherwise the

graph is all black.) It may be convenient to view the surface from a particular

viewpoint (x, y, z); in this case, after Lighting->True, a comma then

ViewPoint->{x,y,z} is added.

The amount of time Mathematica takes to solve a system of differential

equations increases as the order of the system increases. For this reason, it is

convenient to reduce the order of the system by introducing the auxiliary

variables p = u’ and q = v’. This yields the system

u ′ = p
v' = q

 −1
 p ′ = 2
(x uu • x u p 2 + 2x uv • x u pq + x vv • x u q 2 )
 xu
 −1
q ′ = 2
( )
x uu • x v p 2 + 2x uv • x v pq + x vv • x v q 2
 xv
(12)
 11

Now, these equations are inputted into Mathematica, as in Figure 1. To make

this easier, the partial derivatives of x are designated as functions of x before

inputting the differential equations.

Figure 1

As an example of the output, geo is calculated for a sphere in Figure 2. Above

the black line is the input; below is the output.

Figure 2

Mathematica is able to solve the geodesic equations numerically by evaluating at

a sequence of points and interpolating to estimate the behavior between the

points. In the following code in Figure 3, the variables {u0, v0} represent the

starting point of the geodesic; ang gives the initial direction of the geodesic,

which combined with the fact that the geodesic is unit-speed, gives the initial
 12

velocity. Therefore, {u0, v0} and ang give the necessary initial conditions to

give a unique geodesic.

Figure 3

Take[..., {a,b}] selects the ath through bth elements of a list and outputs

them in another list; First[...] gives the first element of a list as a number.

So First[Take[geo..., {3,3}] gives the numeric value of u’’.

To display the solution on a surface, two different plots are made and

saved. First, the equations are solved: solution1=solvetest1[

x,0,{u0,v0},ang,tfin]. Next, the geodesic is plotted:

g1=ParametricPlot3D[Evaluate[Append[x[u[t],

v[t]], AbsoluteThickness[3]] /. Solution1],

{t,0,tfin}],

and the surface is plotted:

g2=ParametricPlot3D[x[u,v], {u, u1, u2}, {v, v1,

v2}, Lighting->True].

Lastly, the two plots are shown together: Show[g1,g2,Lighting->

True,ViewPoint->{x,y,z}]. Finding the proper viewpoint depends on the

 13

geodesic, and often requires trial and error. Now, we give specific examples of

this procedure by finding and plotting geodesics on several surfaces.

The Sphere

For the patch for the sphere, we see that G = xu ● xu = r2, and E = xv ● xv

= r2cos2u, so the patch for the sphere is a v-Clairaut patch. So, using a unit

E −k2 cos 2 v − k 2 k
sphere for simplicity, v ′ = = 2
and u ′ = . So
EG cos v cos 2 v

k k tan v
u=∫ = sin −1 ( ) + c . This implies
cos v cos 2 v − k 2 1− k 2

k k tan v 1− k 2
u=∫ = sin −1 ( ) + c , and so sin(u − c) = tan v ⇒
cos v cos 2 v − k 2 1− k 2 k

1− k 2 1− k 2
sin(u − c) cos v = sin v ⇒ (sin u cos c − cos u sin c )cos v = sin v ⇒
k k

cos c 1 − k 2 sin c 1 − k 2
(13) sin u cos v − cos u cos v = sin v
k k

Note that the components of the patch for the sphere are here in this last

equation. Setting x = cosu cosv, y= sinu cosv, z = sinv, we see that equation

(13) implies that the points all lie on the same plane. So a geodesic on a sphere

is the intersection of the sphere and a plane.

Moreover, note that if a point (x, y, z) on the sphere satisfies (13) so too

does (-x, -y, -z). So if the plane in question includes (x, y, z), it includes (-x, -y, -

z). Hence it includes the line between them, and therefore the plane passes

through the origin. So a geodesic on the sphere is the intersection of the sphere

and a plane passing through the origin—a great circle! Choosing two points that
 14

are not antipodal—that is, not on opposite sides of the sphere—allows one to

solve for a unique k and c.

Using the Mathematica procedure above verifies the fact that the

geodesics on a sphere are great circles. Figure 4 shows a few geodesics on a

sphere.

Figure 4

Returning to the example of a plane flying from New York City to Madrid,

we can now plot its path and find the distance that it will travel. Madrid is located

at 40û 26ý = 40.43û North latitude and 3û 42ý = 3.70û West longitude; New York

City is at 40û 47ý = 40.78û North latitude and 73û 58ý = 73.97û West longitude,

according to [5]. The patch given for the sphere is called the “geographical

patch” because the point x(u,v), u and v give the latitude and longitude when

converted into degrees. u gives longitude; a positive value is East longitude, and

a negative one is West longitude. v gives latitude; a negative value is South

latitude. This means that New York = x(-73.97 π/180, 40.78 π/180), and Madrid

= x(-3.70 π/180, 40.43 π/180). Since the geodesic segment between them is an
 15

arc of a great circle, the length of this arc is rθ, where r is the radius of the earth

and θ is the angle between the vector from the center of the earth to New York

and the vector from the center of the earth to Madrid. Using Mathematica we can

find θ by

theta = ArcCos[(sphere[-73.97 Pi/180, 40.78

Pi/180].sphere[-3.70 Pi/180, 40.43 Pi/180])/(r^2)]

This gives θ = 0.904385. The radius of the earth can be estimated as 6378 km

(3963 miles). So the distance from New York to Madrid along the shortest path

is 5768.17 km (3589.50 miles). The radius of the circle around the fortieth

parallel is 2
x NYC + y NYC
2
− z NYC
2
= 2443.48 km; the center of this circle is (0, 0,

4165.83). So the angle between the vector from the center of this circle to New

York and the vector from the center of this circle to Madrid (assuming for

simplicity both are on the 40th parallel) can be found in Mathematica as

angle=ArcCos[(earth[-73.97 Pi/180, 40 Pi/180]-

{0,0,4165.83}).(earth[-3.70 Pi/180, 40 Pi/180]-

{0,0,4165.83}))/(rho^2)], where rho is the radius of the 40th parallel.

angle = 1.22644, so the distance from New York to Madrid along the 40th parallel

is 5992.2 km. So the great circle method is significantly shorter.

The Cylinder

The cylinder can be parameterized by x(u, v) = (r cos(u), r sin(u), v).

xu●xu = r2, xu●xv = 0, and xv●xv = 1, so this is clearly a Clairaut patch. Solving

the Clairaut integral for v in terms of u gives v = ku. Now, unit-speed implies 1 =
 16

t kt
r2 u’2 + v’2 = (r2 + k2) u’2. So u (t ) = + c ⇒ v(t ) = + c , for some
r2 + k2 r2 + k2

constants k and c. So a geodesic α on a cylinder has the form

  t   t  kt 
α (t ) =  r cos + c , r sin  + c , + c  . This is the equation
 
 r +k   r +k  r +k
2 2 2 2 2 2
 

of a helix. Two interesting degenerate cases are where the initial velocity is

horizontal (k=0)—in which case the geodesic is a circle going around the

cylinder—and where the initial velocity is vertical (k=∞)—in which case the

geodesic is a vertical line up the side of the cylinder. These cases and one non-

degenerate helical case are shown, using the Mathematica procedure, in Figure

5.

Figure 5
 17

The Torus of Revolution

The torus of revolution is the surface formed by revolving a circle in the y-z

plane, which does not touch the z-axis, about the z-axis. The surface looks like a

doughnut. A patch for the torus of revolution is x(u, v) = ((a + rcosv) cosu, (a +

rcosv) sinu, rsinv). r here represents the radius of the circle being revolved; a

represents the distance from the circle to the z-axis. The case r=1, a=2 is shown

in Figure 6.

Figure 6

The patch here is a v-Clairaut patch. In fact, all surfaces of revolution admit a

Clairaut patch. If the curve being revolved is (g(v), h(v)), the patch is x(u,v) =

(h(v)cosu, h(v)sinu, g(v)). In the case of the torus, E = xu • xu = (a + rcosv)2, and

G= xv • xv = r2. Yet the Clairaut integral cannot be solved explicitly. Here is

where the implicit approach of the Mathematica procedure is useful. A few

examples may be instructive. If the initial direction is π/2, the geodesic goes

around the “doughnut hole” as in Figure 7

 18

Figure 7

If the initial point is above the largest circle around the outside of the torus, and

the initial direction is 0 (i.e. horizontal), then the geodesic turns downward to the

bottom of the torus, then turns upward to the top, then downward, and so on.

Figure 8 is an example.

Figure 8

The “Two-Banana” Surface

As we saw with the torus of revolution, there are cases were with a

Clairaut patch no explicit equations for the geodesics can be found. If the patch

we are working with for a surface is not Clairaut, it becomes even less likely that

a search for explicit geodesics will be successful. For example, consider the

surface parameterized by x[u,v]=((2+cos(u))cos(v), cos(v)sin(u), sin(v)). We call

 19

this surface the “Two-Banana” surface because it looks like two bananas meeting

at their ends. This surface is shown in Figure 9.

Figure 9

For this patch, G= xv • xv = cos2v + (5 + 4cosu)sin2v, so the patch is not Clairaut

or v-Clairaut. We can use the Mathematica procedure geo to get an idea of what

the geodesic equations look like in this case. The results, in Figure 10, are not

pretty.

Figure 10

Mathematica, however, can still graph the geodesics implicitly. If the geodesic

begins horizontally along the widest circle of a banana, it will go around that

circle as in Figure 11. If it begins vertically along the outermost part of a banana,

it will loop around both bananas as in Figure 12. In general, however, the

geodesics stay on one banana and have complicated shapes. For example,

Figure 13 shows the geodesic starting at u=0, v=0 and with initial direction π/4.
 20

Figure 11

Figure 12

Figure 13
 21

The Hyperbolic Plane

Thus far all the surfaces examined have been surfaces in R3. For

surfaces not in R3 the procedure requires modification. For example, we

consider the hyperbolic plane. This is modeled as the upper half-plane U2, the

set of all (u, v) such that v>0. The properties of the plane, however, are different

from those of R2. Here, the inner product depends on where it is applied. If two

vectors w1 and w2 are applied at (u, v), w1◦w2=(w1●w2)/v2, where ◦ is the inner

product in U2, and ● is the standard dot-product in R2. The upper half-plane with

this inner product is called the Poincaré plane. So x(u,v)= (u,v), and xu(u,v) =

(1,0), but xu(u,v) ● xu(u,v) = 1/v2, and xv(u,v) = (0, 1/v), but xv(u,v) ● xv(u,v) =

1/v2. So this is a v-Clairaut patch, and so

2
k k2 1  c
u=∫ dv ⇒ ku = ∫ dv ⇒ ku = −v 2 − k 2 + c ⇒ k 2  u −  = 1 − k 2 v 2 ⇒
1 1 v  k
2
−k2 −k 2

v v2
2
1  c
= v 2 +  u −  . This gives the equation of a circle with radius 1/k centered at
 k
2
k

(c/k, 0). Since we are dealing with the Poincaré plane, the geodesics are the

upper halves of these circles. The initial conditions determine k and c. In the

case k=0, the half-circle degenerates to a vertical half-line.

Given two points (u1, v1) and (u2, v2) in the Poincaré plane, if v1 = v2, then

they are connected by a vertical half-line but by no circle. If v1 ≠ v2, then the only

geodesic passing through (u1, v1) and (u2, v2) is the half-circle centered at

 u 22 − u12 + v 22 − v12 
 ,0  with radius v2
−
(u 2
1 − 2u1u 2 + u 22 − v12 + v 22 )
2

. This geodesic
2(u 2 − u1 ) 4(u 2 − u1 )
1 2
 

can be graphed in Mathematica using the procedure shown in Figure 14.

 22

Figure 14
For example, Figure 15 shows the geodesic in the Poincaré half-plane passing

through (1,2) and (3,4).

Figure 15

Conclusion

The concept of “line” is a very intuitive and fundamental concept in our

everyday lives. While the concept of “geodesic” is less intuitive on many

surfaces, it is no less fundamental. For this reason, it is critical to know what the

geodesics look like on a surface in order to gain an intuitive understanding of the

geometry of a surface. Mathematica is an invaluable aid in this process, as

Mathematica bypasses the need to perform laborious calculations in solving,

implicitly or explicitly, the geodesic equations.

 23

References

[1] Braun, M. Differential Equations and Their Applications. New York:

Springer-Verlag, 1983. p. 289.

[2] Gray, H. Modern Differential Geometry of Curves and Surfaces with

Mathematica. Second edition. Boca Raton: CRC Press, 1998. p. 613,

873-874, 939, 947, 960, 966-967.

[3] O’Neill, B. Elementary Differential Geometry. Second edition. San Diego:

Academic, 1997. p. 98, 148, 320, 375, 418-419.

[4] Oprea, J. Differential Geometry and Its Applications. Upper Saddle River,

NJ: Prentis Hall, 1997. 155, 160-162.