Curves in Four-Dimensional Space*

Lee Mun Yew

Raffles Junior College

§l Introduction to Curves

1.1 Definition of a Curve

A curve in a space of any dimension can be seen as the trip taken by a

moving particle. Its position at a particular instant can be determined
parametrically by expressing the co<?rdinates of the point as functions of
time. Therefore, the position, {3, of the particle in three dimensions is
given as follows:
f3(t) = (f3I(t),f32(t),f33(t)).
Time, t, is defined on an open interval on the real lineR. The func-
tions {31 , {32 , and {33 must be differentiable so that the curve will have no
corners like A, B, C as in Figure 1.1. This also means that the particle
must be moving at all times; for otherwise, situations as in Figure 1.1 may
anse.
B

Figure 1.1

*Revised version of a report submitted in connection with the Science Enrichment Programme
jointly organized by the Faculty of Science, National University of Singapore and the Ministry of Edu-
cation, Singapore.

72
 Some examples of curves in three dimensions are :-
(1) Straight line : f3(t) = (p1 + q1 t, p2 + q2 t, p 3 + q3t) where Qi =j:. 0 and
P1, P2, P3, Q1, Q2, q3 are constants.
(2) Circle: f3(t) = (acost,asint,O). This is a circle on the x-y plane
with radius a and centre (0, 0, 0).
(3) Circular helix: f3(t) = (acost,asint,bt). (See Figure 1.2.)

2.

,..--
-L

Figure 1.2

At each particular instant, there is a velocity vector representing the

velocity of the particle. This vector is always tangent to the curve and is
non-zero at all times. Let's call this vector T:

T(t) = d/3 = ( d/31 d/32 d/33) .

dt dt ' dt ' dt

73
1.2 Description of a Curve

In examining a curve, what is important is the shape and not the speed
of the particle moving along the curve. Therefore, given any curve we can
force the particle to move at a constant speed of one unit per second, i.e.,
ITI = 1. By reparametrizing its coordinates, we turn (3 from a function of
time, t, into a function of distance along the curve, s. dT Ids now measures
the rate of change of direction of T but not its magnitude.
In Figure 1.3, at point A, JdT I dsJ is small but at B, JdT I dsJ is large.
Therefore, JdT I dsj should be a good measure of curvature. Kappa, K., is
defined as

""=~~~I= ~~~I·
So K. is a function -of s.

Figure 1.3

For the circle (J(s) = (coss,sins,O),

K. =I~~I= V 2
cos s + sin s
2
=
1.

Since K. is a measure of curvature, this implies that a circle has constant

curvature which is what we would expect. In curves of -h igher dimensions,
there are other properties besides "'-· This suggests that we can find other
functions like K. which give a quantitative description of the curve. The
Frenet formulae show us how other functions are obtained. The aim of
this project is to examine the Frenet formulae in four dimensions.

74
§2 The Frenet Formulae

2.1 The Frenet Formulae in Two Dimensions

A two-dimensional curve (i.e., contained in a plane) has curvature r;, at a
particular position. If T is the tangent of the curve at that point, then
T · T = 1 since the magnitude of T is one. Differentiating with respect to
s, we have
dT
2 ds · T = 0.
This implies dT Ids is perpendicular to T unless dT Ids is zero. Let r;, be
the magnitude of dT Ids, and N its unit vector. Then
dT
ds = ~tN. (1)
We have T · N = 0 since N is perpendicular to T. Differentiating, we
obtain
dT dN
-·N+-·T=O. (2)
ds ds
From (1), taking dot product with N, and since N · N = 1, we get
dT
- · N = r;,N · N = r;,.
ds
Substituting into (2), we find that
dN
ds . T = -r;,.

Similarly, N · N = 1 implies
dN
-·N=O.
ds · .
Now any vector x can be expressed as (x · y)y + (x · z)z where y and
z are perpendicular unit vectors. Therefore,

~ · T) T + ( ~ · N) N = -r;,T.
dN
ds (

So we get
dT dN
ds = ~tN, -ds = -r;,T'
which are the Frenet Formulae in two dimensions.

75
2.2 The Frenet Formulae in Three Dimensions

We now derive the Frenet formulae in three dimensions using a method

which can be generalized to higher dimensions.
As before N is the unit vector of dT Ids and K is the magnitude of
dT Ids. Now we define a unit vector B which is perpendicular to T and
N. By taking dot product with T and N we see that the vector

dN _ ( dN .
ds ds
r) T _ ( dN . N) N
ds

is perpendicular to T and N. As a result, we can let this vector be r B

where r is a scalar known as the torsion. By the usual method, we find
that T · dN/ds = -K, N · dNids = 0. Thus,

dN
ds = -KT + rB.

In three dimensions, any vector can be expressed as the sum of three

perpendicular vectors. We can let

dB
ds = aT + bN + cB.
By taking dot product with T, N, B and differentiating B · T = 0, B · W =
0, B · B = 0, we find that a = 0, b = -r, c = 0. Thus the Frenet formulae
are
dT dN dB
ds = KN, -ds = -KT+rB ' ds = -rN.

2.3 The Frenet Formulae in Four Dimensions

To extend into four dimensions, we follow the same procedure till we reach
dNIds = - KT + r B. By the same principle, we let D be a unit vector
perpendicular toT, Nand B. Then for some scalar u,

u D = dB -
ds
(r . dB) T -
ds
(N .dB) N -
ds
(B .dB) B.
ds

76
Differentiating T · B = 0, N ·B = 0, B ·B = 1, we get
dT dB dB dB
ds · B + ds · T = (KN) · B + ds · T = 0, i.e., -·T=O·
ds '
dN dB dB dB
ds · B + ds · N = (-KT + r B)· B + ds · N = 0, t.e., -·N=-r·
ds '
dB
and 2 ds · B = 0.
Therefore
dB
ds = -rN +aD.

Let
dD
ds = aT + bN + cB + dD.
By taking dot product of both sides with T, N, B and D and differenti-
ating T · D = 0, N · D = 0, B · D = 0, D · D = 1, we get a = b = d = 0,
c = -a. Thus the Frenet formulae in four dimensions are:

dT dN dB dD
ds = KN, -ds = -KT+rB ' -
ds
· = -rN+aD
' ds = -aB.

From the above, we can see that the Frenet formulae for higher di-
mensions are likely to have the same general pattern. In fact, it can be
proved by the same procedure that in each dimension, the same pattern
will occur.

§3 Properties of K, r and a

3.1 Properties of a

In O'Neill's "Elementary Differential Geometry," it is shown that a three-

dimensional curve with K > 0 is a plane curve if and only if r = 0. This
theorem suggests that if a curve in four dimensions has K > 0, r 1- 0, then
the curve is confined to three-dimensional space if and only if a = 0.

77
Proof Let f3 be a curve confined to three dimensions. Then, there exist
vectors x andy such that ((3(8)- x) · y = 0 for all 8. Differentiation yields
(3' (8) · y = 0 or T · y = 0. Differentiating again, we get dT I d8 · y = 0. So by
the Frenet formulae (~tN) · y = 0, or N · y = 0. Differentiating once more,
we have dN / d8 · y = 0. By the Frenet formulae, ( -~tT + r B) · y = 0. But
T · y = 0, so B · y = 0.
T·y = 0, N ·y = 0, B·y = 0 imply that y is perpendicular toT, Nand
B. Hence D can be written as ± : • Differentiating, we get dDid8 = 0.
1 1
dDjd8 = uB by the Frenet formulae. Therefore, uB = 0. Now B i= 0, so
this implies that u = 0.
Conversely, if u = 0, then dD I d8 = -u B = 0. This means that D
must be a constant. Define a function f such that /(8) = ((3(8) - (3(0)) · D
for all 8. Then
is = (3'(8) · D =T ·D = 0.
Thus, /(8) has the same value for all 8. Taking 8 = 0,

/(0) = ((3(0) - (3(0)) · D = 0.

Therefore, ({3(8)- (3(0)) · D = 0 for all 8. From here, we conclude that the
curve f3 must lie entirely in the three-dimensional space orthogonal to the
vector D.

3.2 The Cylindrical Helix

The cylindrical helix is defined as a curve for which there exists a fixed
unit vector u such that T · u is constant along the curve. By the definit~on
of dot product,
T .u = IT II u Icos 0 = cos 0

where 0 is the angle between T and u. Thus 0 is a constant.

In O'Neill's "Elementary Differential Geometry," it is shown that rl~t
is constant along a three-dimensional helix. We will now try to find the
corresponding properties of a four-dimensional helix. We have T·u =cos 0.
Since 0 is a constant,
dT du
-·u+-·T=O.
d8 d8

78
Now dujds = 0, so dT jds · u = 0. From the Frenet formulae we get
K-N · u = 0. Assuming that "' =j:. 0, we have N · u = 0, and

u = (u · T)T + (u · N)N + (u · B)B + (u · D)D

=(cos O)T + aB + 1D

where a = u · B and 1 = u · D. Thus

1 = Iu I = y'cos 2 (} + a2 + 12 ,
2
I.e., a 2 + 1 2 = sin 0.

This suggests that we should define an angle 4> by a = sin(} cos</>,

1 = sin(} sin 4> such that

We then obtain

u = (cosO)T + (sinOcos<f>)B + (sin0sin¢)D.

Since u is a constant, differentiating and using the Frenet formulae again,

we find

0 = (cos O)K-N - ~~(sin 4> sin 0) B + (sin(} cos¢)( -r N + uD)

+ ~~{sin(} cos 4> )D + (sin(} sin 4> )( -u B)

= ("'-cos(}- r sin(} cos <f>)N- sinO sin¢ ( ~~ + u) B

+sin(} cos 4> ( ~~ + u) D.
When a sum of non-parallel vectors is zero, the individual components
are zero. Therefore,

"' cos (} - r sin (} cos 4> = 0 {3)

sin (}sin 4> ( ~~ + u) = 0 {4)

sin(} cos 4> ( ~~ + u) = 0 (5)

79
But sin fJ is a constant. If sin fJ = 0, then (J = 0 or 7r. So T = ±u. Thus
T is a straight line in the direction of u. This case is not of interest. So
from (4) we have
d</>
sin¢= 0 or ds +u = 0.

By a similar reasoning, from equation (5) we get

d</>
cos 4> = 0 or ds +u = 0.
But sin</> and cos</> cannot be equal to zero at the same time, so d</> Ids+ u
must be zero. This yields d</> Ids = -u. Since we are interested in four-
dimensional curves, so u ¥- 0, which means that </> is not a constant. From
(3),
cos (J
• (J A. = cot fJ sec </>.
sm cos 'f'
Now cot (J is a constant but not sec</>. Thus r I K. is not constant.
The significance of the above discussion is that the result obtained is
different from the situation in the thi:ee-dimensional case where a cylin-
drical helix always has r I K. as a constant.
In the previous sections, we saw that the four-dimensional Frenet
formulae and their properties resemble those of the three-dimensional for-
mulae. The above result shows, however, that we cannot always guess the
behaviour of the four-dimensional case from the three dimensional case.

case

Figure 3.1

80
 By the way, the angle¢ does have some geometric significance. Since
u is fixed, as we move along the curve, the vectors T, N, B and D will
move about u in a certain way. If we neglect the vector N and fix the
frame T, Band D, we will get Figure 3.1. From the figure, we see that¢
represents the angle as shown.

§4 Simplest Curves

4.1 Simplest Curves in Two Dimensions

Apart from the straight line which is only one-dimensional, the circle is one
of the simplest two-dimensional curves. The circle also has K, as constant.
Therefore, we can define a simplest curve in two dimensions to be one in
which K, is constant.
We will now investigate whether there are other two dimensional sim-
plest curves by using the Frenet formulae :
dT dN
- = K,N, - = -K,T.
ds ds
Since K, is constant, differentiating the first expression, we obtain
2
d T = K, dN = -K,2 T.
ds 2 ds
By the theory of differential equations,
T = WsinAs +X cos As
where W and X are any constant vectors. Comparing with d2 T / ds 2
-K,2 T, we see that A= K,. The curve .B(s) becomes

.B(s) =
I W
Tds = -TcosAs
X .
+ TsmAs +C.
Although W, X and C are arbitrary constant vectors, they must satisfy
ITI = 1 because we assumed this in our derivation of the Frenet formulae
(refer to Section 2). Therefore, the only possible value for W and X are
(1,0) and (0,1) respectively. In this case

.B(s) = ( -~cosAs, ~sinAs) +C

which clearly represents a circle of radius 1/ A= 1/ K, with C as centre.

81
4.2 Simplest curves in Three Dimensions

A simplest curve in three dimensions has " and r as non-zero constants.

The Frenet formulae in three dimensions are

dT dN dB
ds = ICN, -ds = -ICT+rB, - =-rN.
ds
Differentiating the first expression, we have

Differentiating once more,

This is similar to the two-dimensional case. If we let A2 = ~C2 + r2 , then

dT .
ds = WsmAs +X cos As.
So,
T = - ~ cos AS + ~ sin AS + Y.
Integrating this, we get

{3 (s) = - ~ sin AS - ~ cos AS + Y s + Z,

where W, X, Y, Z are any three-dimensional vectors which satisfy ITI = 1.
The circular helix (Figure 1.2 is obtained by letting W, X, Y and
Z to be (1,0,0), (0,1,0), (0,0, 1-1/A2 ) and (0,0,0) respectively. We
then have

{J(s) = ( - : 2 sin As,- A~ cos As, V(1- 1/ A

2
) s) .

82
4.3 Simplest Curves in Four-dimensions

A simplest curve in four dimensions should be one with K., r and u constant
and non-zero. The Frenet formulae in four dimensions are

dT dN dB dD
ds = ~tN, -ds = -~tT+rB ' -ds = -rN +uD, -
ds
= -uB.

We differentiate the first equation continuously until we get an equa-

tion involving the vector T only.

d?T dN
--
2
= ~t- = -~t 2 T + ~trB. (6)
ds ds
d 3T 2 dT dB
--
ds3
= -K.-
ds
+~er-
ds
= - K2 dT (
. - +~tr -rN +uD
)
ds
dT dT
= -K.2 - - r2 - + ~eruD.
ds ds

(from(6))

(7)

Let T = V cos >.s +W sin >.s where V and W are arbitrary constant
vectors. Then

d2 T
ds 2 = ->. 2 (V cos >.s + W sin >.s) = ->. 2 T, (8)

d4 T = - >. 2 ( - >. 2 T) = >. 4 T (9)

ds 4

Substituting (8) and (9) into (7), we have

83
Since T =f 0, we have the following quartic polynomial equation

(10)
which can be reduced -to a quadratic equation in ..\2 • Solving, we have

The discriminant

since K,, r, u =f 0. Hence ). 2 has two real and unequal values. Furthermore

/\,2 + r2 + u2 > y'(/\,2 + r2 + u2)2 _ 4K,2U2.

Hence ..\ 2 has two real, unequal and positive values. ). can therefore take
four values, ..\ 0 , -Ao, ..\1, -..\1.
When .A= -Ao,

T = V cos( -.A 0 s) + W sin( -.A 0 s)

= V cos A0 s + (-W) sin A0 s.

Since V and W are arbitrary, the solution is similar to that where ). = ..\0 •
Therefore, we need only consider two values of .A : ..\0 and ..\ 1 •
Let
T 0 = V cos A0 s + W sin.A 0 s,
T1 = X cos A1 s + Y sin ..\ 1 s.
Since T 0 and T1 are two solutions of the differential equation (7), we get

Thus T0 + T1 is also a solution of the differential equation. By the theory

of differential equations, there are no other solutions.
Integrating T0 + T1 , we have

84
 Take V = ~(1,0,0,0), W = ~(0,1,0,0), X= ~(0,0,1,0),
Y = ~(0,0,0,1), Z = (0,0,0,0), and we get the following curve:

With this choice of V, W, X, Y and Z, Tis automatically a unit vector.

4.4 Properties of the Four-dimensional Simplest Curve

In this section, we examine some of the more interesting properties of the

curve described in equation (12). A~ and A~ are the two solutions for A2
in equation (10). Now A~ +A~ is the sum of roots and A~ A~ is the product
of roots. Hence,

I.BI =

-21 ( -1 -r2u2
u2 + r;,2
1 )
- +r;,2
- •

Since r;,, r and u are all constant, we see that the curve lies on a four-
dimensional sphere. When r;, or u increases, then I.BI decreases, hence the
sphere is sma:ller. However when r increases, the sphere becomes larger.
When r;, = r = u, I.BI = ;~.
Although it is impossible for us to have a true picture of what the
curve is actually like in four dimensions, we can try to have a glimpse by
examining the curve with one dimension 'cut off'. Since ,B(s) is given by
(12), cutting off any dimension will produce the same type of curve. Let's
cut off the last dimension, to obtain

85
 A rough sketch of the curve is shown in Figure 4.1 resembling a spring
with both ends packed together. The derivation of this sketch is as follows.
The x and y coordinates correspond to a circle. The z coordinate is a sine
function which means that a particle moving along the curve travels faster
in the middle than at the ends.

Figure 4.1

An interesting question is whether the original curve repeats itself.

For it to repeat itself, there must exist a constant a such that

Comparing the first component, we have

, _ {(2m+ l)1r- .A0 s- >.oa (13)

AoS-
2m7r + >. 0 s + >. 0 a. (14)

Comparing the second component, we have

(15)
>. { 2n7r- >.os- .A 0 a
2n7r + >.os + .A 0 a. (16)
08
=

86
where m, n are integers. (13) and (15) cannot be true for all s. So (14)
and (16) must be true which means that -.A 0 a = 2m1r. Similarly, by
comparing the third and the fourth components, we get -.A 1 a = 2n1r.
Thus, .A 0 /.A 1 = mjn. Since m and n are integers, the value .A0 /.A 1 must
be a rational number in order for the curve to repeat itself. Referring to
(11), we find that

"'2 + r2 + u2 ± J(K2 + r2 + u2)2 _ 4~e2u2

"'2 + r2 + u2 =f y'(K2 + r2 + u2)2 _ 4K2u2

We can see that it is unlikely that ). 0 / .A 1 will come out to be a rational

value. This means that the chance of the curve repeating itself is very
small.
In our three-dimensional world, the four-dimensional Frenet formulae
may seem irrelevant and useless. However, this is not so. The Frenet
formulae have been applied in many areas of Physics. This is possible
because we can let one of the axes in four dimensions to be time. Thus,
this three-dimensional time graph shows the motion of a particle in three
dimensions. The Frenet formulae can therefore be used to describe this
motion. In "Classical and Quantum Gravity," Vol. 5, No. 7, there is a
research paper written by B R Iyer and C V Vishveshwara whose title
is 'The Frenet-Serret Formalism and Black Holes in Higher Dimensions'.
The paper examines the motion of charged particles in homogenous elec-
tromagnetic fields using the Frenet formulae for dimensions higher than
three.

References
[1] Barrett O'Neill, Elementary Differential Geometry, Academic Press,
1969.
[2] B R Iyer & C V Vishveshwara, The Frenet-Serret Formalism and
Black Holes in Higher Dimensions, Classical and Quantum Gravity 5
(7), 1988.
[3] Eisenhart Luther Pfahler, An Introduction to Differential Geometry,
Princeton University Press, 1947.

87