You are on page 1of 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225328790

Hyperbolic Calculus

Article  in  Advances in Applied Clifford Algebras · June 1998


DOI: 10.1007/BF03041929

CITATIONS READS

48 1,053

2 authors, including:

Adilson E Motter
Northwestern University
225 PUBLICATIONS   12,052 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Asymmetry-induced synchronization View project

Chimeras as cluster sync patterns View project

All content following this page was uploaded by Adilson E Motter on 04 April 2014.

The user has requested enhancement of the downloaded file.


HYPERBOLIC CALCULUS
A. E. Motter1,2 and M. A. F. Rosa1
1
Department of Applied Mathematics - IMECC
State University at Campinas (UNICAMP)
CP 6065, 13081-970, Campinas, SP, Brazil
2
e-mail motter@ime.unicamp.br

(Received: March 2, 1998, Accepted: March 28, 1997)

Abstract. The complex numbers are naturally related to rotations and dilatations
in the plane. In this paper we present the function theory associate to the (universal)
Clifford algebra for IR1,0 [1], the so called hyperbolic numbers [2,3,4], which can be
related to Lorentz transformations and dilatations in the two dimensional Minkowski
space-time. After some brief algebraic interpretations (part 1), we present a “Hyper-
bolic Calculus” analogous to the “Calculus of one Complex Variable”. The hyperbolic
Cauchy-Riemann conditions, hyperbolic derivatives and hyperbolic integrals are in-
troduced on parts 2 and 3. Then special emphasis is given in parts 4 and 5 to con-
formal hyperbolic transformations which preserve the wave equation, and hyperbolic
Riemann surfaces which are naturally associated to classical string motions.

1. Introduction
The geometrical interpretation of the complex numbers as the plane of Argand-
Gauss together with the natural relation of the complex numbers with the
Euclidean structure of the plane are the main reasons for the power of Complex
Calculus.
We can fix a complex number z = a + ib and define Tz : C I →C I given by
I ≡ IR2 , the plane with the Euclidean metric in the
Tz (w) = zw. If we identify C
2
 Tz :IR →
usual way, the application Tz can be seen as a linear transformation
a −b
IR2 . This one can be identified with the matrix Tz = Ta+ib = . We
b a
note that the ring   
a −b
TCI = ; a, b ∈ IR
b a

Advances in Applied Clifford Algebras 8 No. 1, 109-128 (1998)


110 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

under matrix addition and matrix product is isomorphic to the ring of the
complex numbers. As transformations of the plane, the (matrix) product of
elements of TC I corresponds to the composition of transformations. If we ex-
clude the null matrix (det(Ta+ib ) = a2 + b2 6= 0) from TC I ∗ , a group
I we get TC
of transformations of the plane. Therefore if we identify the complex numbers
with the points of the plane, we identify the complex numbers acting as linear
transformations in the plane (by complex product) with the ring TC, I obtained
from the group TC I ∗ of rotations and dilatations plus null matrix.
Now we turn to the hyperbolic numbers IP = {t + hx; t, x ∈ IR} with sum
and product defined similarly to the complex case, but with h2 = 1 (instead
of i2 = −1). It has been useful to identify the hyperbolic numbers with the
2-dimensional Minkowski space-time (see [2]). This because, as the complex
numbers are naturally related to the Euclidean structure of the plane, the
hyperbolic numbers are to the Lorentzian structure of such Minkowski space.
Again we define the hyperbolic conjugate of w = t + hx by w̄ = t − hx and
||w||2M ≡ ww̄ = t2 − x2 is its “squared Minkowski modulus” (w is time-like if
||w||2M > 0, light-like if ||w||2M = 0 and space-like if ||w||2M < 0).
This definition shows that the hyperbolic numbers are the universal Clifford
algebra for IR1,0 and isomorphic to 2IR = IR ⊕ IR, the direct sum of two real
numbers algebras [1]. It’s well known that IP = 2IR together with IR, C, I IH
(quaternions) and 2IH = IH ⊕ IH form the fundamental blocks employed in
the classification of Clifford algebras (very well presented in [5]). Despite the
algebraic simplicity the existence of zero divisors makes that the hyperbolic
calculus here presented cannot be included in the general scheme of usual
Clifford analysis (that found in [6] for example). Therefore we are dealing with
a crucial point, that will be lighted on parts 2, 3 and 4. Now we follow with a
geometric interpretation similar to that given to complex numbers.
We fix the hyperbolic number z = a + hb ∈ IP to define Tz : IP → IP by
Tz (w) = zw. If we identify IP = IM 2 , the Minkowski 2-dimensional space time,
by
 
t
w = t + hx ≡ ∈ IM 2 ,
x
2 2
 Tz : IM → IM ,
the application Tz can be seen as a linear transformation
a b
which can be identified with the matrix Tz = Ta+hb = . Furthermore,
b a
the ring
  
a b
T IP = ; a, b ∈ IR
b a
Advances in Applied Clifford Algebras 8, No. 1 (1998) 111

under matrix addition and matrix product is isomorphic to the ring of the
hyperbolic numbers.
 Then if we consider z = hφ, a “pure hyperbolic number”,

0 φ
we have Tz = and
φ 0
∞  n  
X 1 0 φ cosh φ sinh φ
exp Tz = = .
n! φ 0
n=0
sinh φ cosh φ

Since exp Tz = Texp z we have exp hφ = cosh φ + hsenhφ for a pure hyperbolic
number.
Now we consider the geometrical meaning of the transformations Tz :
IM 2 → IM 2 of the 2-dimensional Minkowski space time generated by hyper-
bolic product with a fixed hyperbolic z. We divide it in three cases:

Case 1: z is time-like, and z = a + hb with a2 − b2 > 0. We define ρ =


(a2 − b2 )1/2 and ε = sgn(a) (+1 if a > 0, −1 if a < 0), then we get
   
a b cosh φ sinh φ
Tz = Ta+hb = = ερ = Tερ exp hφ ,
b a sinh φ cosh φ
where φ is such that cosh φ = εa/ρ, sinh φ = εb/ρ. Then z = ερ exp hφ (polar
form of a time-like hyperbolic) and the above decomposition shows that the
hyperbolic product by a fixed time-like hyperbolic corresponds to a proper ho-
mogeneous Lorentz transformation (ε exp hφ) followed by a dilatation (ρ).

Case 2: z is space-like, and z = a + hb with a2 − b2 < 0. We note that


z = h(b + ha) where the hyperbolic b + ha is time-like. If we define ρ =
(b2 − a2 )1/2 , ε = sgn(b) and φ by cosh φ = εb/ρ, sinh φ = εa/ρ we get

b + ha = ερ exp hφ and z = hερ exp hφ.


This decomposition shows that the product by a fixed space-like hyperbolic is a
proper homogeneous Lorentz transformation (ε exp hφ) followed by a dilatation
(ρ) and a “permutation of space and time” (h). This because the transforma-
tion Th : IM 2 → IM 2 interchanges space and time, Th (t + hx) = x + ht. It can
be interpreted as a superluminal Lorentz transformation (see [2,7]).

Case 3: Now z is light-like, that is z = a + hb with a2 − b2 = 0. Then, since


det Tz = a2 − b2 = 0, Tz is singular (but with rank one, except in the case
112 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

z = 0).

The set {z ∈ IM 2 s.t. ||z||2M = 0} of all light-like hyperbolic numbers is the


“light cone” of IM 2 and is the union of the two straight lines x = ±t in the
plane IM 2 . Note that these lines contain the image of Tz : IM 2 → IM 2 for z
light-like.
From the above interpretation we see that the group of transformations of
IM 2 defined by
  
a b
T IP ∗ = ; a, b ∈ IR, a2 − b2 6= 0
b a
is generated by the proper homogeneous Lorentz transformations, dilatations
and the space time permutation. It has been obtained from the ring T IP by
excluding the light-like hyperbolic, which are divisors of the zero as hyperbolic
numbers, and (singular) projections on the light cone as linear transformations.
We also note that T IP ∗ plus the hyperbolic conjugation generate the linear
conformal group of IM 2 .
The simplest ring of complex functions f : C I →C I is the ring of the linear
function. These functions are compositions of elements of TC I and translations.
The “Calculus of one Complex Variable” deals with the “holomorphic func-
tions” and its properties. We then say that f : C I →C I is holomorphic at a
point z0 ∈ C I if a certain limit (the complex derivative) exists (see [8]). And we
can think the holomorphic functions as “locally linear functions”. That can be
approximated by a linear function in a small neighborhood of a point.
Analogously the simplest ring of hyperbolic functions f : IP → IP are the
linear ones in the sense of hyperbolic product and addition. These are compo-
sitions of element of T IP and translations. And it seems natural to characterize
the “hyperbolic holomorphic functions” as being functions f : IP → IP with a
good linear (in the hyperbolic product sense) approximation,
f (z) − f (z0 ) ∼= f 0 (z0 )(z − z0 )
or ∆f ∼ = Tf 0 (z ) (z − z0 ),
0

in a small neighborhood of a point. This is our starting point to develop a


“Calculus of one Hyperbolic Variable” by analogy with the Complex one.

2. Derivative
In the complex case the Cauchy-Riemann conditions for f at P0 have a nice
geometric interpretation. They are equivalent to say that Jacobian matrix of
Advances in Applied Clifford Algebras 8, No. 1 (1998) 113

f at P0 is an element of TC. I This characterizes the holomorphic functions


as exactly those functions which have a good linear approximation close to a
point (this means differentiable) in such a way that this linear approximation
can be written as a linear complex function (this comes from Cauchy-Riemann
conditions).
Now we turn to the hyperbolic numbers, IP = IM 2 . We employ in IM 2
the topological structure of IR2 , this could be seen as contradictory and is
criticized (see [9]). But it is the convention adopted in all the literature (see
[10,11] for example) and we assume here this simplified point of view (leaving
the suggestion made in [9] for a posterior work). This means that, despite of
the Lorentzian structure of IM 2 we will be using the concept of neighborhood
and making limits as if we were in IR2 .
Let f : IP → IP, f (t + hx) = u(t, x) + hv(t, x) be differentiable when seen
as a function of IM 2 to IM 2 (or from IR2 to IR2 ). Then we say that f satisfies
the hyperbolic Cauchy-Riemann condition if
∂u ∂v ∂u ∂v
= and = .
∂t ∂x ∂x ∂t
Therefore the Jacobian matrix of f belongs to T IP, Df ∈ T IP since
   
∂u/∂t ∂u/∂x a b
Df = = ∈ T IP,
∂v/∂t ∂v/∂x b a
with a = ∂u/∂t, b = ∂v/∂t. And the linear transformation Df corresponds to
the hyperbolic product with the fixed hyperbolic a + hb = ∂u/∂t + h∂v/∂t.
It can be the composition of proper Lorentz transformations with dilatations
when ∂u/∂t + h∂v/∂t is time-like, also followed by a a space-time permutation
when is space-like. Or a projection in the light cone when ∂u/∂t + h∂v/∂t is
light-like.
The defining equation for the differential of f ,
     
du ∂u/∂t(t0 , x0 ) ∂u/∂x(t0 , x0 ) ∆t
df = = ,
dv ∂v/∂t(t0 , x0 ) ∂v/∂x(t0 , x0 ) ∆x
that approximates ∆f = f (t0 + ∆t, x0 + ∆x) − f (t0 , x0 ) close to (t0 , x0 ) ∈ IM 2
can be written in hyperbolic form
 
∂u ∂v
df ≡ du + hdv = (t0 , x0 ) + h (t0 , x0 ) (∆t + h∆x)
∂t ∂t | {z }
≡dz
114 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

when f satisfies the hyperbolic Cauchy-Riemann conditions.


In this case f is approximated by a linear hyperbolic transformation. It
seems natural to define a hyperbolic holomorphic function as a differentiable
function satisfying hyperbolic Cauchy Riemann conditions. But we exaggerate
a little bit.

Definition 1. Let f : IP → IP, f (z) = u(t, x) + hv(t, x) for z = t + hx,


we say that f is hyperbolic holomorphic, or that it has hyperbolic deriva-
tive at z0 = t0 + hx0 , if u and v are C ∞ at (t0 , x0 ) and satisfy the hy-
perbolic Cauchy-Riemann conditions ∂u/∂t = ∂v/∂x and ∂u/∂x = ∂v/∂t
at this point. Then we say that f 0 (z0 ) = ∂u/∂t(t0 , x0 ) + h∂v/∂t(t0 , x0 ) =
∂v/∂x(t0 , x0 ) + h∂u/∂x(t0 , x0 ) is the hyperbolic derivative of f at z0 .

Well, this is not so exaggerated, since in the complex case holomorphism


implies C ∞ . Furthermore we want to keep the result that the hyperbolic deriva-
tive of a hyperbolic holomorphic function is hyperbolic holomorphic. In the
complex case this is done by employing the Cauchy Integral Formula, which
we will not have in its full strenght for the Hyperbolic Calculus (see part 3).
Therefore, with the above definition we have that f 0 as a function is C ∞
since its components ∂u/∂t and ∂v/∂t are derivatives of C ∞ functions. Then
we have from Schwarz’ rule that

Proposition 1. If f : IP → IP is hyperbolic holomorphic in an open set U ⊆


IP , its derivative is also hyperbolic holomorphic as a function f 0 : U ⊆ IP → IP .

The “chain rule” gives us that the hyperbolic derivative of the composite
is the hyperbolic product of the hyperbolic derivatives,

Proposition 2. If f : IP → IP is hyperbolic holomorphic and also is g : IP →


IP , the composite gof : IP → IP is hyperbolic holomorphic and

gof 0 (z0 ) = g 0 (f (z0 ))f 0 (z0 ), for z0 ∈ IP.

Now we observe that, if our hyperbolic holomorphic functions are locally


linear, it should be expected that the hyperbolic product of two hyperbolic
holomorphic functions be hyperbolic holomorphic, since hyperbolic product is
closed in the ring of hyperbolic linear functions.

Proposition 3. If f, g : IP → IP are hyperbolic holomorphic functions also


are f + g and f · g defined by hyperbolic addition and product. Furthermore
Advances in Applied Clifford Algebras 8, No. 1 (1998) 115

(f + g)0 (z0 ) = f 0 (z0 ) + g 0 (z0 )


(f · g)0 (z0 ) = f 0 (z0 )g(z0 ) + f (z0 )g 0 (z0 ).

And then

Proposition 4. The hyperbolic polynomials are hyperbolic holomorphic func-


tions and if

f (z) = z m , f 0 (z) = mz m−1 (0 < m ∈ ZZ).


1
If z ∈ IP is not light-like we can define f (z) = = z/||z||2M . Therefore
z
f (z) = 1/z is hyperbolic holomorphic outside the light cone, then f 0 (z) =
−1/z 2 . From this and from propositions 2 and 3, we have

Proposition 5. If g(z) is not light-like, and f, g are hyperbolic holomorphic


f 0 (z)g(z) − g 0 (z)f (z)
functions, f /g is also holomorphic and we have (f /g)0 (z) = .
(g(z))2
The elementary transcendental complex functions are defined from the com-
plex exponential. We have defined the hyperbolic exponential by employing the
isomorphism between the rings IP and T IP . From above it is given by

f (z) = exp z = et (cosh x + h senh x),

and follows from Df that f (z) = ez is hyperbolic holomorphic with f 0 (z) = ez .


Since IP and T IP are isomorphic and detDf (P0 ) = ||f 0 (z0 )||2M for a hyper-
bolic holomorphic function (z0 ≡ P0 ), from the inverse function theorem, we
have

Proposition 6. If f : IP → IP is hyperbolic holomorphic and f 0 (z0 ) is not


light-like, there are open sets V, W ⊆ IP with f 0 (z) not light-like for z ∈ W
and z0 ∈ W and there is a function g : V → W such that f (g(w)) = w.
Furthermore g is hyperbolic holomorphic and

g 0 (w) = 1/f 0 (z) = f 0 (z)/||f 0 (z)||2M ; z = g(w)


(hyperbolic inversion) for w ∈ V .
116 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

In the case of the hyperbolic exponential ||f 0 (z)||2M = e2t 6= 0 and the above
proposition holds a local inverse function. We define the hyperbolic logarithm
g : V ⊆ IP → IP with

V = {u + hv ∈ IP s.t u2 − v 2 > 0, u > 0}, and


g(u + hv) = log(u2 − v 2 )1/2 + h argIP (u + hv).

We define the hyperbolic argument of u + hv by x = argIP (u + hv) such that


sinhx = (u2 −vv2 )1/2 . From Dg we see that g is a hyperbolic holomorphic function
and g 0 (z) = 1/z in its domain. The exponential function is a bijection between
IP and the open set V = {u + hv ∈ IP s.t. u2 − v 2 > 0, u > 0}, the future
pointing time-like hyperbolic, and the above defined logarithm is its inverse.
We now remember a result due to [12]. There are functions which satisfy
definition 1, hence hyperbolic holomorphic, but not analytic (in the real sense).
One example is the function given by exp(−1/z 2 ) outside the light cone and
by its limit in the light cone. This is the first indication that this structure
requires special attention. In the Clifford analysis developed in [6] for example,
where there is a restriction to the Euclidean case, the monogenic function (that
is, the holomorphic functions) are necessarily analytic.

3. The Integral
The existence of complex derivative is the condition for local integrability, and
the complex integral is an antiderivation for holomorphic functions. These two
facts can be generalized to the hyperbolic case if we give the natural definition:

Definition 2. f = u + hv, f : IP → IP is a hyperbolic continuous function and


Γ = Imγ is a Jordan arc with γ : [a, b] → IP and γ 0 (τ ) is continuous except for
a finite number of points. We define the hyperbolic integral of f on Γ as the
line integral
Z Z Z
f dz = udt + vdx + h vdt + udx.
Γ Γ Γ
Z
The dependence of f dz only on the final and initial points γ(b) and γ(a)
Γ
for a holomorphic f follows from the following analogous of Morera’s theorem.
Advances in Applied Clifford Algebras 8, No. 1 (1998) 117

Proposition 7. Let f : IP → IP be hyperbolic holomorphic and a closed


Jordan arc Γ, then
I
f dz = 0.
Γ
For this we apply Green’s theorem and the Cauchy-Riemann conditions for
f.
It is not hard to see that the hyperbolic integral has the usual properties
of a line integral and that the above proposition implies

Z f : D ⊆ IP → IP holomorphic in the simply connected


Proposition 8. Let
open set D. Then f dz depends only on the final and initial points of the
Γ
Jordan arc Γ.

And the existence of the hyperbolic


Z derivative Zis the condition for local
integrability. Since the integrals udt + vdx and vdt + udx depend only
Γ Γ
on the extreme points, we can find scalar potentials U (t, x) and V (t, x) for
the vector fields (t, x) → (u(t, x), v(t, x)) and (t, x) → (v(t, x), u(t, x)). Then
we define the hyperbolic holomorphic function F = U + hV , it has hyperbolic
derivative F 0 (z) = f (z) and the Fundamental Theorem of Calculus gives us

Proposition 9. If FZ0 (z) = f (z) is holomorphic and Γ is contained in the do-


main of F , we have f dz = F (γ(b)) − F (γ(a)), that is, the hyperbolic line
Γ
integral works as hyperbolic antiderivative.

The function f (z) = 1/z is not defined in the light cone but if Γ is the
Jordan arc of (with initial point zi and final point
Z zf ), contained in the region
τ+ = {t + hx ∈ IP, t2 − x2 > 0, t > 0}, we have 1/zdz = log zf − log zi . In
Γ
the case zi and zf belong to the hyperbole t − x = ρ2 , employing the polar
2 2

form,
Z
1/zdz = h(φf − φi ),
Γ

which is a local analogue of complex relation z1 = 2πi, the difference is that


H
the hyperbolic angle has an infinite range. In fact the above relation can be
118 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

extended by (we put φi = −φf = φ)


Z
1 1
lim dz = 1,
φ→∞ 2hφ Γ(φ) z

where Γ(φ) is any curve in τ+ between ρ exp(−hφ) and ρ exp hφ. Unfortunately
we have no results giving more general Cauchy Integral Formulae. We think
that the existence of holomorphic functions that are not analytic may be re-
lated with the lack of an integral formula in this theory. In Clifford analysis,
particularly in complex analysis, these ideas are very linked.
From the geometric point of view the problem is that the codimension of the
light cone is one and it is not possible to “avoid the light cone” in the sense that
there is no curve involving the origin together with a non divergent kernel to
the integral. This observation seems obvious but it plays a fundamental role in
higher dimensional function theories on Clifford algebras. Taking biquaternions
I 2 ) as an example, the algebra has eight real dimensions
(Clifford algebra for C
and the light cone only six. Then the codimension two of the cone (inside
the algebra) permits us to find an integral formula for biquaternions [13,14].
Therefore in this point of view we are treating a special case.
However there is a ”Cauchy Algebraic Formula” in this case (we refer [12]
for a full discussion). Here we only state this very important result. For this
we define the elements

1 1
α= (1 + h), α= (1 − h),
2 2

which are an idempotent basis in IP (α2 = α, α2 = α and αα = 0).

Proposition 10. Let f : D ⊆ IP → IP holomorphic in the simply connected


open set D. For z ∈ D, and z1 , z2 ∈ IP so that z + αz1 , z + αz2 ∈ D, we have

f (z) = αf (z + αz1 ) + αf (z + αz2 ).

Follows from this that if f is holomorphic for all z, it is uniquely determined


everywhere by its values on the hyperbole t2 − x2 = ρ2 . It’s again analogous to
the complex one: a complex holomorphic function inside the circle is uniquely
determined by its values in the boundary.
Advances in Applied Clifford Algebras 8, No. 1 (1998) 119

4. Conformal Transformations
From now on all the functions are C ∞ , and we consider f : C
I →C
I holomorphic
( ∂f
∂z = 0). It is well known that the two component functions are harmonic, we
write ∆f = 0, and that the operators ∂/∂z and ∂/∂z factorizethe Laplacian,
∂ ∂ ∂ ∂ 1
= = ∆.
∂z ∂z ∂z ∂z 4
This permits us to apply the complexified version of Euler’s method to the
equation ∆f = 0 obtaining the general solution for the Laplace equation as
a combination of pure holomorphic and antiholomorphic parts, and it is a
generalization of D’Alambert’s solution to the wave equation, where the general
solution is a sum of waves travelling in opposite directions.
Now, if f = u + hv is hyperbolic holomorphic, from the hyperbolic Cauchy-
Riemann conditions, the two component functions satisfy the wave equation.
We write 2f = 0. In the complex case the conjugate harmonic functions have
orthogonal level curves. In the hyperbolic case we have two conjugate solutions
of the wave equations with orthogonal level curves in the Minkowski sense, since
∂u ∂v ∂u ∂v
− = 0. As pointed in [3], we can write u and v in D’Alambert
∂t ∂t ∂x ∂x
form,

u(t, x) = Φu (t − x) + Ψu (t + x),
v(t, x) = Φv (t − x) + Ψv (t + x),

and from Cauchy-Riemann we get Ψu = Ψv and Φu = −Φv (up to constants).


That is, the wave described by v is obtained from u by changing the signal of
the part propagating in the positive x direction.

We can also make the hyperbolic change of variables (t, x) 7→ (z, z) with
z = t + hx, z = t − hx, and define the operators
   
∂ 1 ∂ ∂ ∂ 1 ∂ ∂
= +h , = −h .
∂z 2 ∂t ∂x ∂z 2 ∂t ∂x
Then the hyperbolic holomorphic functions are those satisfying ∂f /∂z = 0.
We say that f is hyperbolic antiholomorphic when its conjugate is holomor-
phic. And since holomorphism is a local condition for integration, in a simply
connected domain we can get the holomorphic primitive F (z) of the hyper-
bolic holomorphic f (z), satisfying F 0 (z) = f (z). Also in the hyperbolic case
120 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

functions of z can be integrated by the usual rules giving another function of


z (the same for z). Again,
∂ ∂ ∂ ∂ 1
= = 2.
∂z ∂z ∂z ∂z 4
and we have a factorization of the hyperbolic wave equation different from
D’Alambert’s one. And the hyperbolic solutions of this equations will be com-
binations of hyperbolic holomorphic and antiholomorphic parts.
In [6] the nonogenic functions satisfy Laplace’s equation (in the correspond-
ing dimension) and analyticity can be seen as consequence of this fact in context
of harmonic analysis. Of course we cannot employ this argument here.
The relation between hyperbolic holomorphism and wave equation is deeper.
So we also generalize the conformal transformation to this case, where the wave
equation will replace Laplace’s one. For that we first define

Definition 3. Let T : U ⊆ IP → IP (U open), we say that T is hyperbolic


conformal if it is hyperbolic holomorphic and inversible. That is, there exists
a hyperbolic holomorphic function T −1 : V → U (V ⊆ IP is obviously open)
such that T ◦ T −1 = Id (identity function).

We note that T and T −1 do not have light-like


 hyperbolic
 derivatives. Then
−1 a b
if z = T (w), w = T (z) we have DT (z) = , with a − b2 6= 0, while
2
b a
   
−1 1 a −b c d
DT (w) = 2 = , with
a − b2 −b a d c
1
c2 − d2 = 6= 0.
a2 − b2
If ge(u, v) is a solution of the wave equation in V ⊆ IP ,

∂ 2 ge ∂ 2 ge
− 2 = 0,
∂u2 ∂v
with g(t, x) = ge ◦ T (t, x), we obtain

∂2g ∂2g
 2 
2 2 ∂ ge ∂e
g
− = (a − b ) − = 0.
∂t2 ∂x2 ∂u2 ∂v 2
Advances in Applied Clifford Algebras 8, No. 1 (1998) 121

From the above we have


Proposition 11. Let T : U ⊆ IP → IP be conformal in U and continuous
in ∂U , then if ge(u, v) is a solution to the wave equation in T (U ) with given
boundary conditions in ∂T (U ), we have the solution g(t, x) = ge ◦ T (t, x) of the
wave equation in U , with the corresponding boundary conditions in ∂U .

Now let’s consider T (z) = log(hz) and T −1 (w) = h exp w as conformal


transformations between the sets U and V as in figure 1. Since t = eu sinh v, x =
eu cosh v the straight lines u = 0 and u = a corresponding to the hyperboles
x2 −t2 = 1 and x2 −t2 = ea , the segment v = 0, 0 ≤ u ≤ a to t = 0, 1 ≤ x ≤ ea .

Fig. 1. Conformal transformation in the Hyperbolic Plane.

Identifying the hyperbolic plane with the two dimensional space time, u is
the space axis and v the time axis. Then the solutions of the wave equation in V
are solutions of a vibrating string problem. These are well known, and through
the conformal transformation we can obtain solutions of the more complicated
problem of solving the wave equation with boundary conditions in U .
In fact it is interesting to solve this problem since (pages 37 to 40 of [15])
it corresponds to solve the wave equation in a rod wich is accelerated rigidly
until the velocity of light. First we consider a material point moving from rest
with constant acceleration. Parametrizing its world line by its proper time we
have s 7→ z(s) = t(s) + hx(s) for the trajectory and z 0 (s), z 00 (s) for velocity
and acceleration respectively. For each s, z 0 (s) is a time-like vector, and since
122 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

the proper time agrees with Minkowski arc lenght we have

hz 0 (s), z 0 (s)iM = z 0 (s)z 0 (s) = 1, (∗)


and the derivation of this quadratic form gives us hz 0 (s), z 00 (s)iM = 0. And
since z 00 (s) is Minkowski orthogonal to z 0 (s) we conclude that z 00 (s) is space-
like.
Therefore we can write

hz 00 (s), z 00 (s)iM = z 00 (s)z 00 (s) = −α2 , (∗∗)


when the material point has a constant proper acceleration.
Now from (∗) and the fact that z 0 (s) is time-like we have z 0 (s) = exp hϕ(s)
for a particle travelling to the future. Then z 00 (s) = hϕ0 (s) exp hϕ(s) and
(∗∗) (ϕ0 (s))2 = α2 , and ϕ0 (s) = α is a constant of the motion (the choice
ϕ0 (s) = −α is analogous). This gives ϕ(s) = αs + β, a linear function of the
proper time, and putting z 0 (0) = 1 + 0h we get β = 0 and z 0 (s) = ehαs .
Integrating we obtain

z(s) = hα−1 exp(hαs) = α−1 sinh(αs) + hα−1 cosh(αs)


if we put the condition z(0) = α−1 . Such z(s) = t(s) + hx(s) with t(s) =
α−1 sinh(αs) and x(s) = α−1 cosh(αs) is the parametrization of the world line
of a particle at rest at t = 0 when its position is x = α−1 , under constant
proper acceleration (figure 2(a)).
This world line obeys the equation x2 − t2 = α−2 , and we consider (figure
2(b)) the motion of a rod in x direction given by the equation x2 − t2 = w2 .
Where for each value of the parameter w we have a point of the rod which is
at x = w when t = 0 and is under a constant proper acceleration given by α =
w−1 . Differentiating for constant w we have 2xdx−2tdt = 0, therefore if we de-
fine the
instantaneous Lorentz factor γ(v) = (1 − v 2 )−1/2 with v = dx/dt we have
γ(v) = x/w. Again differentiating x2 − t2 = w2 , but for fixed t, we have

wdw dw
dx = =
x γ(v)
and we quote [15].
...“Since this applies to any two neighbouring points in the aggregate rep-
resented by x2 − t2 = w2 the role aggregate ‘moves rigidly’, like an unstressed
rod”...
Advances in Applied Clifford Algebras 8, No. 1 (1998) 123

Fig. 2. (a) Point particle under constant proper acceleration. (b) Rigdly accelerated
rod.

And the conformal transformations (see figure 1 and associated discussions)


permit us to obtain solutions of the wave equation in this rigidly acceleration
rod from the solutions of the wave equation for a rod at rest. It is not hard to
see that one of these solutions will be
    
1 t
Ψkw (t, x) = cos k ln(t2 − x2 ) cos w sen h−1 ,
2 (t2 − x2 )1/2
where sinh−1 is the inverse of the hyperbolic sine (this solution cones from the
usual coskx coswt for the rod at rest).

5. Hyperbolic Riemann Surfaces


In the complex case the Riemann Surfaces are special cases of C ∞ 2-dimensional
Manifolds: those whose all coordinate changes are holomorphic functions (em-
ploying the identification IR2 ≡ C).
I The classical example is the Riemann
sphere. Let S 2 be the two dimensional sphere, S 2 = {(x, y, z) ∈ IR3 s.t.
x2 + y 2 + z 2 = 1} with the topology induced from IR3 . Then U1 = S 2 − {N }
and U2 = S 2 −{S} where N = (0, 0, 1) and S = (0, 0, −1) are an open cover for
S 2 (N and S are the north and south poles). Now we employ two coordinate
systems π : U1 = S 2 − {N } → C I and σ : U2 = S 2 − {S} → C I to make S 2 a
124 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

Riemann surface. Then π is the stereographic projection given by


π( sen θ cos ϕ, sen θ sen ϕ, cos θ) = cotg(θ/2)eiϕ ,
(for (x, y, z) ∈ S 2 − {S} with x = sen θ cos ϕ, y = sen θ sen ϕ and z = cos θ)
and σ = π ◦ J, the composite of the stereographic projection with the rotation
J : S 2 → S 2 of π through the x axis. The two coordinate changes, σ ◦ π −1 and
π ◦ σ −1 , are holomorphic functions in their domains since they correspond to
the complex inversion. We note that the stereographic projection through the
south pole instead of σ wouldn’t make S 2 a Riemann surface (but only a C ∞
2-dimensional manifold). For πS ( sen θ cos ϕ, sen θ sen ϕ, cos θ) = tg( θ2 )eiϕ
and we note that we could have defined σ as σ = C ◦ πS where C : C I →C I is
the complex conjugation, C(z) = z̄.
Now we turn to the hyperbolic case. Let M be a C ∞ 2-dimensional mani-
fold, with a countable open cover Uj , j = 1, 2, 3.... We employ the identification
between IR2 and IP such that the coordinate system Φj : Uj ⊆ M → IP are
hyperbolic valued functions. And if it is possible to obtain a refinement of the
open cover such that all the coordinate changes χjk : Φj (Uj ∩ Uk ) ⊆ IP → IP
are hyperbolic holomorphic, we say that M is a hyperbolic Riemann surface.
We note that any coordinate change χjk is hyperbolic conformal since it has
an inverse χkj which is hyperbolic holomorphic.  
a b
We note that the Jacobian of these coordinate changes has the form
b a
with a2 − b2 6= 0, since χjk is inversible its hyperbolic derivative χ0jk =
a + hb cannot be light-like. If we can refine again the open cover such that
all hyperbolic derivatives χ0jk be time-like (a2 − b2 > 0) then the hyperbolic
Riemann surface would be orientable manifold as in complex case.
Our hyperbolic Riemann surface can be made Lorentzian manifold by em-
ploying the metric induced by the coordinate systems and a partition of unity
subordinated to them (see [16]). Then a oriented hyperbolic Riemann surface
would be said space time orientable when we can make a new refinement of
the open cover obtaining only coordinate changes χjk for which a > 0 in
χ0jk = a + hb.
What would be the analogous of the Riemann sphere S 2 = {(x, y, z) ∈ IR3
s.t. x2 + y 2 + z 2 = 1}? A good candidate would be the one sheet revolution
hyperboloid which we denote by
S 1,1 = {(t, x, y) ∈ IR1,2 s.t x2 + y 2 − t2 = 1}.
It is a 2-dimensional surface in the 3-dimensional Minkowski space IR1,2 (one
time and two space dimensions).
Advances in Applied Clifford Algebras 8, No. 1 (1998) 125

The surface S 1,1 can be thought as the set of world lines of an expanding
closed circular string. Consider the intersection of S 1,1 with the (x, t) plane,
from the discussion on part 4, this intersection is the world line of a particle
under constant proper acceleration in the x direction. By taking other intersec-
tions we see that the string is “expanding” in such a way all its points have the
same radial proper acceleration. And again from part 4 we observe that if this
string has a nonzero thickness, it is expanding rigidly in the radial direction.
Now let’s make S 1,1 a hyperbolic manifold. Our starting point is that the
stereographic projection π : S 2 − {N } → IR2 was a conformal transformation.
In fact, parametrizing the sphere by (sin θ cos ϕ, sin θ sin ϕ, cos θ) we see that
it has a natural metric induced from IR3 given by dSesf
2
= dθ2 +sin2 θdϕ2 . And
2
in the plane parametrized by (ρ cos ϕ, ρ sin ϕ) we have dSplane = dρ2 + ρ2 dϕ2 .
The stereographic projection applies a point with spherical coordinates (θ, ϕ)
to a point with coordinates (ρ(θ), ϕ) in the plane, where ρ(θ) = cotg(θ/2).
Therefore

1 θ
2
dSplane = dρ2 + ρ2 dϕ2 = cossec4 ( ) {dθ2 + sin2 θ dϕ2 },
|4 {z 2 }
| {z }
dS 2esf
λ

and the stereographic projection is conformal (see [17]) with conformal factor
λ = 1/4 cossec4 (θ/2).
Analogously we can parametrize S 1,1 inside our (t, x, y) space by (sinh Φ, cosh Φ
cos θ, cosh Φ sin θ) and it has a Lorentzian metric induced from the Minkowski
space IR1,2 (see [17]) given by dShip
2
= (cosh2 Φ)dθ2 − dΦ2 . And the Minkowski
2
plane IP has the metric dSplane = dρ2 − ρ2 dϕ2 .
It is not hard to see that a conformal stereographic projection cannot keep
the angle ϕ, that is we cannot make Φ = ϕ. And if we write

2
dShip = cosh2 Φ{ |{z}
dθ2 − sech
| {z
2
ΦdΦ}2 },
dρ2 /ρ2 dϕ2

this gives us a suggestion. Let’s define ϕ = ln{|sechΦ + tgh Φ|} which is a


C ∞ function for Φ > −Φ0 . Where Φ0 is the solution of sech Φ0 = tgh Φ0 , or
exp(Φ0 ) = 2 + exp(−Φ0 ).
We also define ρ = eθ , therefore

2
dShip = ρ−2 cosh2 Φ {dρ2 − ρ2 dϕ2 }.
| {z } | {z }
λ 2
dSplane
126 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa

Fig. 3. Our candidate to Hyperbolic Riemann sphere: ρ varies between eπ/2 and
e5π/2 (because θ is between π/2 and 5π/2) while ϕ varies between −∞ and ln M
(where M is the maximum of sechΦ+ tghΦ).

And the above defined stereographic projection (θ, Φ) → (ρ, ϕ) is conformal


with conformal factor λ = (e−θ cosh Φ)2 .
If we measure θ between π/2 and 5π/2, the above stereographic projection
gives us a coordinate system π1 : U1 ⊆ S 1,1 → IP , with
U1 = {(sinh Φ, cosh Φ cos θ, cosh Φ sin θ) s.t. π/2 < θ < 5π/2 and Φ > −Φ0 },
and

π1 (sinh Φ, cosh Φ cos θ, cosh Φ sin θ) =


eθ exp{h ln |sechΦ + tghΦ|}.
= |{z}
| {z }
ρ ϕ

We have another coordinate system π2 : U2 ⊆ S 1,1 → IP defining π2 by the


same analytical expression employed for π1 , but measuring θ between −π/2
and 3π/2. The coordinate change χ12 : π1 (U1 ∩ U2 ) ⊆ IP → IP is given by
χ12 (z) = eπ z, being holomorphic. And U1 , U2 is a cover for a bit more than
one half of S 1,1 . To cover the whole hyperboloid we add other two coordinate
systems. To define π3 we again measure θ between π/2 and 5π/2. But

π3 (sinh Φ, cosh Φ cos θ, cosh Φ sin θ) =


e−θ exp{−h ln |sech(−Φ) + tgh(−Φ)|},
= |{z}
| {z }
ρ hϕ
Advances in Applied Clifford Algebras 8, No. 1 (1998) 127

therefore U3 = {(sinh Φ, cosh Φ cos θ, cosh Φ sin θ) s.t. π/2 < θ < 5π/2 and
Φ < Φ0 }. And we have that χ13 : π1 (U1 ∩ U3 ) ⊆ IP → IP is holomorphic since
χ13 (z) = 1/z.
The last coordinate system is π4 , with the same analytical expression of π3
but with θ measured between −π/2 and 3π/2. And S 1,1 is our proposal for the
Hyperbolic Riemann sphere. We defy the reader to find a better one.
As the above example shows, the Hyperbolic Riemann surfaces are natu-
rally associated to string motions and deformations since they are Lorentzian
manifolds with one space and one time dimension.

6. Conclusion

We have presented here a “Hyperbolic Calculus” analogous to the “Calculus of


One Complex Variable”. Hyperbolic Cauchy-Riemann conditions, Hyperbolic
derivatives and Integrals, Hyperbolic Conformal Transformations and Riemann
surfaces are obtained. As we have shown, these can have interesting applica-
tions to Special Relativity. Even with the usual topology we have obtained
some differences in relation to the usual Clifford analysis. We were not able to
generalize Cauchy’s Integral Formula, we only have obtained a simplified ver-
sion of it. One of the investigations that this paper suggests is in this direction.
For this we may have to employ the convergence from a different topology for
IP (see [9]).
This work also suggests investigation on other hypercomplex calculus. A
first temptation is to define f : IRp,q → IRp,q (IRp,q is IRp+q with a metric of
signature p, q) as holomorphic when it is C ∞ and locally a linear conformal
transformation. But for this Liouville’s theorem (see [17]) works as a “don’t
go” theorem. Since it shows that the conformal transformations from IRp,q to
IRp,q are linear (what would trivialize our holomorphic functions).
Another option is to define a subring of the C ∞ functions f : V → W as
holomorphic when these preserve in certain sense some interesting algebraic
properties of the vector spaces V and W . For these speculations see the refer-
ences of [18].
A third direction is the study of a Calculus of Many Hyperbolic Variables.
This is an easy generalization of the Calculus of Many Complex Variables and
will be the subject of a posterior work.
128 Hyperbolic Calculus A. E. Motter and M. A. F. Rosa
View publication stats
Acknowledgements
We are grateful to Prof. A. Rigas, Prof. W. A. Rodrigues Jr. and L. G. G. V.
Dias da Silva for useful discussions, and to FAPESP for the financial support.

References
[1] Porteous I. R., “Topological Geometry” Cambridge University Press, Cam-
bridge, 1981.
[2] Fjelstad P., Extending Special Relativity via the Perplex Numbers, Am. J.
Phys. 54 (5) (1986).
[3] Laurentiev M. and B. Chabat, “Effects Hydrodynamiques et Modèles
Mathématiques” Mir, Moscow, 1980.
[4] Assis A. K. T., Perplex Numbers and Quaternions, Int. J. Math. Educ. Tech. 22
(4) (1991); G. Sobczyk, The Hyperbolic Number Plane, preprint; V. Majernik,
Spec. Sci. Technol. 6 (1983) 189.
[5] Keller J., Quaternionic, Complex, Duplex and Real Clifford Algebras, Adv.
Appl. Clifford Alg. 4, No 1 (1994), 1.
[6] Delanghe R., F. Sommen, V. Soucek, F. Brackx and D. Constares, “Clifford Al-
gebra and Spinor-Valued Functions”, Kluwer Academic Publishers, Dordrecht,
1992.
[7] Recami E. and R. Mignani, Lett Nuovo Cimento 4 (1972) 144.
[8] Apostol T. M., “Mathematical Analysis”, Addinson-Wesley Publishing Com-
pany, 1974.
[9] Zeeman E. C., The Topology of Minkowski Space, Topology, 6 (1967) 161.
[10] Sachs R. K., and H. Wu, “General Relativity for Mathematicians”, Springer
Verlag, 1977.
[11] Synge J. L., “Relativity: The General Theory”, North Holland, 1960.
[12] Antonuccio F.,“Semi-Complex Analysis and Mathematical Physics”, preprint.
[13] Imaeda K., A New Formulation of Eletromagnetism, Nuovo Cimento B 32
(1976), 138.
[14] Ryan J., Cells for Harmonicity and Generalized Cauchy Integral Formulae,
Proc. London Math. Soc. 60 (3) (1992), 295.
[15] Rindler W., “Introduction to Special Relativity”, Oxford University Press,
1989.
[16] O’Neill B., “Semi-Riemannian Geometry”, Academic Press, 1983.
[17] Dubrovin B. A., A. T. Fomenko, S. P. Novikov, “Modern Geometry”, v. I
Springer Verlag, 1984.
[18] Gürsey N. and H. C. Tze, Ann. Phys. 128 (1980), 29.

You might also like