History of Mathematics 0038

Lecture Notes 20: The Discovery of Non-Euclidean Geometry

1. Discovery

Carl Gauss

Janos Bolyai

Nikolai Lobachevsky
 2

The failure of repeated attempts to prove the parallel postulate led Carl Gauss (1777-1855), Nikolai
Lobachevsky (1792-1856), and Janos Bolyai (1802-1860) independently to develop a consistent
geometry based on the hypothesis of the acute angle. This involved retaining Euclid’s first four
postulates while discarding the parallel postulate and replacing it with the postulate that through
any point an infinite number of straight lines can be drawn parallel to a given straight line.

Gauss was one of the greatest mathematicians of all time, making fundamental contributions to
number theory, differential geometry and electromagnetism among many other fields. He began to
reflect on the theory of parallels while still an undergraduate. He became convinced that the parallel
postulate could not be proved and decided to explore the possibility of a geometry based on the
HAA which he called non-Euclidean geometry. He derived many of the fundamental properties of
this geometry but he declined to publish his discoveries for fear of ridicule by an ignorant public. He
only communicated a few results to trusted correspondents whom he begged to keep the
information confidential. In 1831, fearing that his ideas might perish with him, he began to write a
memoir in which he wrote down a few theorems on parallels. But he abandoned his memoir when,
to his astonishment, he received a copy of Janos Bolyai’s comprehensive treatise on the subject.

Janos Bolyai, the son of one of Gauss’ university friends, was an infant prodigy. At the age of ten he
played first violin in the local string quartet in the Transylvanian town where his father taught
mathematics. On occasions he even substituted for his father in class. Among his many
accomplishments he was proficient in several languages: Hungarian, German, Romanian, Italian,
French, Latin, Chinese and Tibetan. At fifteen he was sent to the Royal College of Engineers in
Vienna. After completing the seven year course in four years he was commissioned into the
engineering corps of the Austrian army where he gained a reputation as a duellist. On one occasion
he defeated thirteen cavalry officers one after another, pausing only to play his beloved violin after
every other duel. In the army he devoted much of his spare time to mathematics, in particular the
theory of parallels, ignoring his father’s advice to stay clear of a subject which had ‘extinguished all
light and joy’ in his own life. Bolyai started out, like so many of his predecessors, with the aim of
proving the parallel postulate. But meeting the inevitable impasse, he decided to explore the
possibility of a geometry in which the postulate is denied. In 1823 he wrote to his father, ‘I have
created a new world out of nothing’. His father advised him to publish his discoveries as soon as
possible since an idea, when the time is ripe, tends to spring up everywhere ‘like violets in early
spring’. Bolyai published ‘The Absolute Science of Space as an appendix to his father’s textbook in
1831. His father sent a copy to his old friend Gauss. Gauss replied that to praise the work would be
to praise himself since Bolyai’s results ‘coincided almost entirely’ with his own meditations on the
subject over the past thirty-five years (see Appendix 1). Bolyai felt utterly deflated. He even
suspected his father of having revealed his ideas to Gauss who was now dishonestly claiming
priority. As for Gauss, though he described Bolyai as ‘a genius of the first rank’, he never sought to
further the career of his old friend’s son, as he might easily have done.
 3

Bolyai failed to obtain recognition for his work from the authorities in Vienna, which was hardly a
surprise given the challenging style of his treatise which even Gauss found difficulty in
understanding. In 1833 he took early retirement from the army and lived the rest of his life on a
small pension in Transylvania. Although he never published more than the 24 pages of his Appendix,
he left 20 000 pages of mathematical manuscripts.

But unknown to both Gauss and Bolyai, in Kazan, a town several hundred miles to the east of
Moscow, a professor of mathematics, Nikolai Lobachevsky, had also discovered non-Euclidean
geometry. Lobachevsky spent virtually his whole adult life at the University of Kazan, first as an
undergraduate, then as a professor and finally as rector. Unfortunately in 1846 he was stripped of
his position without explanation and he lived out the rest of his life in reduced circumstances. His
ideas were first published in Russian in 1829 under the title On the Principles of Geometry but his
treatise was rejected when it was submitted to the St Petersburg Academy of Sciences. In 1840 he
published Geometrical Investigations on the Theory of Parallel Lines, written in German with the
hope of attracting a wider audience. This brought his work to the attention of both Gauss and Bolyai.
In a letter to a friend Gauss said that he had found nothing really new in Lobachevsky’s work but
praised the mastery of his exposition. He recommended Lobachevsky for membership of the
Scientific Society of Göttingen (which is more than he did for Bolyai). Bolyai, for his part, was
astonished at the similarities between Lobachevky’s work and his own and described Lobachevsky’s
treatise as a masterly achievement.

2. Neutral Definition of Parallel Lines

Gauss, Bolyai and Lobachevsky all provided a neutral definition of parallel lines, that is to say one
that neither assumed nor ruled out the validity of the parallel postulate. Gauss and Bolyai gave the
following definition:

When the straight line AM does not cut the straight line BN but every other straight line passing
through A within MAB (such as AR) does cut BN, then AM is said to be parallel to BN. This
definition does not exclude the possibility that there may be other straight lines passing through A
outside MAB (such as AP) which also fail to meet BN. Gauss proved three fundamental properties
of parallel lines: 1) a straight line retains the property of parallelism along all its points (e.g. if AM is
parallel to BN, then GZ will also cut BN 2) parallelism is reciprocal (that is, if AM is parallel to BN,
then BN is parallel to AM e.g. BF will cut AM) 3) parallelism is transitive (that is, if AM and BN are
both parallel to a third line, then they are parallel to each another.
 4

3. Neutral Construction of the Parallel

Bolyai also proved a neutral construction using a straight edge and compass to draw the parallel to
a given straight line through any point:

In the figure AN is the given straight line and D is the given point. The construction of the parallel
DM may be obtained as follows. First draw DB perpendicular to AN. Next take any point A on AN and
draw AE perpendicular to AN (where E is yet to be determined). Next draw DE perpendicular to AE
(which determines E). Next with centre A and radius AO equal to ED, draw the circle meeting DB at
O, creating angle AOB = z. Finally draw angle BDM = z. Then DM will be parallel to AN, as required.

This construction is valid regardless of whether the parallel postulate is assumed to be true or not. If
it is assumed to be true, then ABDE will be a rectangle, so AO = ED = AB. Consequently the point O
will coincide with B, which implies that angle AOB and therefore angle BDM is a right angle, as one
would expect. If on the other hand the parallel postulate is rejected, with the hypothesis of the right
angle being replaced by the hypothesis of the acute angle, then ABDE will be a Lambert quadrilateral
with three right angles and an acute angle at D. In this case ED will be greater than AB. Consequently
AO, being equal to ED and greater than AB, will meet DB at a point O between D and B, creating an
acute angle AOB = z. Thus angle BDM = z, called the angle of parallelism corresponding to the
segment BD between the parallels DM and AN (see Section 7 below), will be acute.

4. Neutral Geometry

Before proceeding to create their new geometry, Gauss, Lobachevsky and Bolyai began by proving a
number of theorems whose validity is independent of the parallel postulate. Perhaps the most
astonishing of these ‘absolute’ or neutral theorems is Bolyai’s three dimensional version of the
parallel postulate itself. Bolyai proved that if the sum of the dihedral angles which two planes
make with a given plane is less than two right angles, the lines common to the planes being
parallel, then the two planes will intersect. This theorem is valid regardless of whether the parallel
postulate is assumed to be true or not:
 5

In the figure BN is drawn parallel to AM. This implies that any line drawn inside MAB such as AC
will cut BN and similarly any line drawn inside ABN will cut AM. MAB is a right angle but no
assumption is made about the magnitude of ABN which will either be a right angle or an acute
angle according to whether the parallel postulate is accepted or rejected. AC is drawn perpendicular
to BN. (If the parallel postulate is assumed to be true, then ABN will be a right angle and the points
B and C will coincide. If ABN is acute, then C will lie somewhere along BN, as shown in the figure).

At this point Bolyai introduced a third dimension. He drew planes PAM and BND, where PAM is
perpendicular to plane ABNM but where the dihedral angle between plane BND and plane ABNM is
acute. In other words the sum of the dihedral angles which the planes PAM and BND make with the
base plane ABNM is less than two right angles. His aim was to show that the planes PAM and BND
meet.

To prove this he added three further constructions. He drew CE perpendicular to BN in the plane
BND. Next he drew AF perpendicular to CE meeting CE at F. (Since the dihedral angle between planes
BND and ABNM is assumed to be acute, the point F will lie somewhere along CE, as shown). This
creates triangle ACF (shaded) Finally he joined B and F. This creates a fourth plane, PABF.

Bolyai’s argument then runs as follows. Suppose that the plane PABF is rotated about AB until it lies
along ABNM. Then AP will lie along AM. What about AF? AC, being perpendicular to BN, represents
the shortest distance between A and BN. But AF < AC, since AC is the hypotenuse of triangle ACF, so
after rotation AF will not reach BN. This in turn implies that BF will lie inside ABN. In this position
BF must intersect AP since AM and BN are parallel. Therefore BF must intersect AP in the original
position too, before rotation. But AP lies in plane MAP while BF lies in plane BND. Hence the two
planes MAP and BND intersect, as required.

5. Hyperbolic Non- Euclidean Geometry

Unable to prove the parallel postulate, Gauss, Lobachevsky and Bolyai independently created a non-
Euclidean geometry, now known as hyperbolic geometry, in which Euclid’s first four postulates are
retained but where the parallel postulate is replaced by a different postulate. The new postulate
states that given a straight line and a point not on the line, more than one straight line can be drawn
through the point parallel to the given line. The figure below illustrates the hyperbolic plane:

BC is the given straight line and A is a point not on the line. AD is perpendicular to BC and AE is
perpendicular to AD. By Euclid’s last neutral theorem (Elements, 1, 27) AE is parallel to BC, the
alternate angles BDA, DAE being equal. But in the hyperbolic plane there will be more than one
straight line passing through A parallel to BC. And if there is one such line, like Ak, it is easy to show
that there must be an infinite number of non-intersecting lines. At the same time there are clearly
an infinite number of lines passing through A, including AD, which intersect BC. It follows that there
must be a line Ah which separates the non-intersecting from the intersecting lines. This line is
defined in hyperbolic geometry as the right hand parallel, meeting the given line at an infinitely
distant point. The angle DAh (marked) is defined as the angle of parallelism corresponding to the
segment DA. By the same token there is also a parallel to the left. It can be shown that the angle of
parallelism is the same on both sides. It can also be shown that lines such as AE and BC which have a
common perpendicular, namely AD, actually diverge in both directions (see Proclus’ attempted proof
of the parallel postulate).

6. The Hypothesis of the Acute Angle

It can be shown that hyperbolic geometry, as defined above, implies the hypothesis of the acute
angle. Thus the sum of the angles of a triangle in hyperbolic geometry is less than two right angles,
the sum of the angles of a quadrilateral is less than four right angles and generally the sum of the
angles of a convex polygon of n sides is less than 2n – 4 right angles, which is the Euclidean formula.
Furthermore the defect of a polygon increases with its size, a fact which implies, as Gauss
recognised, the existence of a maximum triangle whose angle sum is zero. Another corollary is that
squares, in the sense of quadrilaterals with four fight angles, do not exist in hyperbolic geometry,
making the calculation of area in terms of unit squares problematical to say the least.
 7

7. The Angle of Parallelism

The angle of parallelism in hyperbolic geometry is defined as the (acute) angle which the parallel
makes at the given point with the perpendicular drawn from the point to the given line. An
important theorem stated by Lobachevsky is that if the perpendicular distance between the point
and the given line is denoted by p, the angle of parallelism  (p) tends to zero as p increases
without bound whereas it tends to a right angle as p becomes infinitely small:

In the figure A is the given straight line (where  represents an infinitely distinct point) and EA is
perpendicular to A. It is easy to show that the angle of parallelism  EA = (p) corresponding to
the distance EA = p is smaller than the angle of parallelism CA = (q) corresponding to the
distance CA = q given that p > q. For in hyperbolic geometry the sum of the interior angles between
parallel lines must be less than two right angles, so EC + CE < 2R. On the other hand
CE + CA = 2R since they are adjacent angles. It follows that EC < CA, as required. It is
clear that as E recedes from A, the corresponding angle of parallelism will tend to zero, while as E
approaches A , the angle of parallelism will approach a right angle.

The fact that the angle of parallelism tends to a right angle (which is always the case in Euclidean
geometry) as the distance p tends to zero, implies that that the theorems of Euclidean geometry
hold true over infinitesimal domains in the hyperbolic non-Euclidean plane. But the precise
relationship between p and the angle of parallelism (p) is yet to be determined (see section 11).

8. The Horocycle

The horocycle is a plane curve peculiar to hyperbolic geometry. (The name was coined by
Lobachevsky and has seemed to stick). Given an array of parallels in a plane meeting at an infinitely
distant point, a horocycle can be defined as the curve which cuts each of the parallels at right angles.
The horocycle therefore can be viewed as a circle of infinite radius. Through any point on any of the
parallels only one horocycle can be drawn:
 8

9. The Linear Constant k

The existence of an absolute unit of distance on the hypothesis of the acute angle was initially
suggested by Lambert. Gauss introduced the notation k to denote this linear constant. Lobachevsky
and Bolyai obtained a geometrical representation of k in terms of concentric horocyclic arcs between
parallels. They showed that if the radial distance separating two horocycles between the same
parallels is x, then the ratio of the larger arc to the smaller arc is given by ex/k, where k is the linear
constant. An immediate corollary of this theorem is that if the radial distance is k itself, then this
ratio is e : 1:

But it should be emphasised that k is not a number but an actual unit of distance, whose value,
unfortunately, cannot be derived mathematically. Its value could only be determined, as Gauss,
Lobachevsky and Bolyai all observed, by some physical experiment, assuming that the geometry of
space is hyperbolic rather than Euclidean. The unit k appears in all the formulae of hyperbolic
geometry.

10. The Horosphere

A key element in the hyperbolic geometry of Bolyai and Lobachevsky is the horosphere. A
horosphere, which can be understood as a sphere of infinite radius, is the surface of revolution
generated by rotating a horocycle about any one of its parallel axes. All such parallel axes are
normal to the surface. The intersection between a plane containing any of the parallel axes and the
horosphere is a horocycle.
 9

The figure shown above depicts the surface of a horosphere EGCF, generated by rotating the
horocycle EC about the axis EM, where M lies at an infinitely distant point. The planes MECP, MEDN
and PCDN, which intersect along the parallel axes Em, CP and DN, form an infinitely long triangular
prism, which might be called a paraprism. This paraprism intersects the horosphere along the
horocycles EC, CD and ED, which form a horocyclic triangle ECD.

The fundamental theorem, proved both by Bolyai and Lobachevky, is that the geometry of the
horosphere is Euclidean. That is to say, if horocycles are substituted for straight lines, the figures
formed on the surface of the horosphere satisfy Euclid’s five postulates. The first four postulates,
being valid both in Euclidean and hyperbolic geometry, will hold true on the horosphere viz: only
one horocycle can be drawn between two points, a horocycle can be infinitely extended, a circle of
any radius can be swept out by a horocycle and all right angles between horocycles are equal on the
horosphere.

What about the parallel postulate? Bolyai’s neutral version of the parallel postulate in three
dimensions (see Section 4 above) implies that if the sum of the dihedral angles which two planes
such as MECP and MEDN make with a third plane PCDN is less than two right angles, then MECP and
MEDN must meet along ME. But since the axes CP and DN are normal to the horosphere, the
dihedral angles which the planes MECP and MEDN make with PCDN are equal to the angles ECD and
EDC respectively. Now ECD and EDC can be regarded as the interior angles formed on the
horosphere when the horocycle DC intersects the horocycles EC and ED. So if the sum of these
angles is less than two right angles, CE and DE, lying respectively in planes MECP and MEDN, must
meet at E. This implies that the parallel postulate and therefore the whole of Euclidean geometry is
valid on the horosphere. So, for example, the sum of the angles of the horocyclic triangle ECD must
be two right angles. Similarly the other formulas of plane trigonometry apply to horocyclic triangles
on the horosphere. All the other Euclidean formulas hold true as well. To take another important
example, just as the area of a circle in Euclidean geometry is given by r2, the area of the circular
domain EGCF swept out on the surface of the horosphere by EC will be equal to  EC2.

11. Hyperbolic Trigonometry

The theorems of plane trigonometry are entirely dependent on the parallel postulate since they are
ultimately derived from Euclid’s propositions on similar triangles. This presents an apparently
insuperable problem in describing the relationship between the sides and angles of triangles in the
hyperbolic plane given that hyperbolic geometry is based on the assumption that the parallel
postulate is invalid.

Bolyai and Lobachevsky overcame this difficulty by employing the Euclidean properties of the
horosphere to derive the trigonometry of the hyperbolic plane. That is to say, they used the familiar
trigonometric ratios of a horocyclic triangle such as ECD in the previous figure to deduce the
relationship between the sides and angles of a hyperbolic plane triangle such as ABC inscribed within
the same paraprism.

It is worth showing how Bolyai deduced the non-Euclidean version of the sine rule for triangles in the
hyperbolic plane:

In the figure angle ABC = R and AM is perpendicular to plane ABC. Then plane ABC is perpendicular
to plane AMBN. Furthermore, given that CB is perpendicular to BA, planes PCDBN, MAEDBN are
perpendicular to each other. It follows, since angle EDC is equal to the dihedral angles between
these two planes, that angle EDC = angle ABC = R. A similar argument implies that
angle CED = angle CAB.

Then the Euclidean sine rule, which is valid on the horosphere, that the sides of a triangle are in the
same ratio as the sines of their opposite angles, implies that
 arc EC : arc DC = 1 : sin angle CED

= 1 : sin ABC

Now in Euclidean geometry circles are in the same ratio as their radii, so

arc EC : arc DC = circle with radius EC : circle with radius DC

10

But the circle swept out by arc EC as it rotates about EM and the circle generated by AC rotating
about the same axis are one and the same, namely CFG. And if another horosphere were formed by
rotating arc DC about ND, a similar argument would show that the circle swept out by DC and the
circle generated by BC about the same axis would be the same. Hence

arc EC : arc DC = circle with AC as radius : circle with BC as radius

It follows that

circle with AC as radius : circle with BC as radius = 1 : sin CAB

Thus in any triangle in the hyperbolic plane the circumferences of the circles with radii equal to
the sides of the triangle are in the same ratio as the sines of the opposite angles.

This theorem makes it possible to apply trigonometrical functions like sin, cos and tan to triangles in
the hyperbolic plane. It is the key to deriving the fundamental identity of hyperbolic geometry

tan [½ ( p)] = e p/k,

which relates the angle of parallelism  (p) to its corresponding segment p. This identity, which can
also be written cot (p) = sinh (p/k), uniquely links the circular and hyperbolic functions in one
equation and it can be used to derive the other trigonometric formulas of the hyperbolic plane. One
such formula is the hyperbolic version of Pythagoras’ theorem for a right angled triangle ABC with
the right angle at A:

cosh a/k = cosh b/k . cosh c/k.

This formula transforms into the Pythagorean theorem a2 = b2 + c2 when a, b, c are very small
compared to k, as can be seen by using the expansion cosh x = 1 + x2/2! + … This reinforces the fact
that Euclidean geometry holds true in the hyperbolic plane if k is assumed to be very large.

12. The Equidistant

An equidistant is defined as the locus of points which all have the same perpendicular distance from
a straight line and lie on the same side of it. In Euclidean geometry the equidistant to a straight line
is itself a straight line. But in hyperbolic geometry the equidistant is a curve (see Alhazen’s
attempted proof of the parallel postulate).

11

Bolyai succeeded in rectifying the equidistant curve:

He showed that if s is the length of arc PQ of an equidistant curve, a the length of its projection MN
on the given straight line AA’ and b the perpendicular distance of any point on the equidistant to
AA’, then

s = a cosh (b/k)

At the same time he introduced the idea of an equidistant surface, a surface each point of which lies
at the same perpendicular distance from a base surface. If the area of the base is given by p, the area
q of the equidistant surface lying at a distance b from the base is

q = p cosh2 (b/k)

13. Spherical Trigonometry

An important result, discovered independently by both Bolyai and Lobachevsky, is the fact that
spherical geometry is independent of the parallel postulate. That is to say, the formulae of
spherical trigonometry can be derived on the assumptions of both Euclidean and hyperbolic
geometry. Furthermore these identities transform into the corresponding identities of hyperbolic
trigonometry by replacing the sides a, b, c of a triangle by ia, ib,ic. For example, given that
cos ix = cosh x, the Pythagorean theorem in spherical geometry

cos a/k = cos b/k . cos c/k

transforms, when a, b, c are multiplied by i, into

cosh a/k = cosh b/k . cosh c/k,

which is the Pythagorean theorem in hyperbolic geometry, verifying Lambert’s insight that the
hypothesis of the acute angle holds true on the surface of an imaginary sphere.

14. Area

The representation of area in hyperbolic geometry is problematical given that unit squares, which
are the usual basis for the measurement of area, do not exist if the parallel postulate is abandoned.
Bolyai proved that the area of any (convex) polygon is given by the product of its defect and the
square of the linear constant.

12

Thus the area A of the finite triangle with angles , ,  is given by A = k2 (  ( +  + )]. Suppose
now that the sides of this triangle are extended, as shown, to produce a triangle of maximum size
whose sides are mutually parallel and whose vertices lie at infinitely distant points. The angle sum of
the maximum triangle is zero and so its defect is . It follows that its area is equal to k2.

15. The Circle

One of the most extraordinary features of hyperbolic geometry is that the number  does not
represent the ratio of the circumference of a circle to its diameter. In fact this ratio increases with
the size of the circle. The formula for the circumference C of a circle of radius r is in fact given by
C = 2 k sinh (r/k) while its area A is given by A = 4  k2 sinh2 (r/2k).

These equations, like the other equations of hyperbolic trigonometry, transform into the equations
of Euclidean geometry if the linear constant k is taken as infinitely great, or if, which amounts to the
same thing, the distance r is taken as infinitely small in comparison to the linear constant (a fact
noted by both Bolyai and Lobachevsky):

C = 2πk sinh r/k = 2πk (r/k + r3/ 3!k3 + … )  2πr (as r/k  0)
 A = 4πk2 sinh2 r/2k = 2πk2 (r2/ 2! k2 + r4/4! k4 + …)  πr2 (as r/k  0).

From the mathematical point of view the fact that Euclidean geometry holds true over infinitesimal
distances allows the infinitesimal calculus to be applied to hyperbolic figures. In this way it is
possible to obtain the rectification and quadrature of circles, equidistants and horocycles by
integration. But there are also consequences for physics. If the geometry of space is hyperbolic
rather than Euclidean, the effect may not be observable over distances small in comparison with k,
which is bound to be an astronomically large unit of distance. Even in a hyperbolic world Euclidean
geometry is likely to remain a valid approximation for all practical purposes, which is just as well.

13

Even more extraordinary is Bolyai’s discovery that in hyperbolic geometry the circle can be squared
with compass and straightedge. In Euclidean geometry this is an impossible task because a square
equal in area to a unit circle must have a side of length √. Unfortunately  is not an algebraic
number (i.e. the root of a polynomial equation with rational coefficients), so it cannot be
constructed with compass and straightedge. But in hyperbolic geometry there exists a linear figure,
the maximum triangle, whose area is k2. The fact that the area of a linear figure and the area of the
circle are both functions of  avoids the need to construct a segment of length √. All that is
required to square the circle is 1) to determine the ratio of its area to the area of the maximum
triangle and 2) to construct a ‘square’ (that is, an equilateral quadrilateral with four equal acute
angles) whose area lies in the same ratio to the area of the triangle. Bolyai provided a method of
constructing such a square using compass and straightedge.

An important element in Bolyai’s construction is the fact that, associated with every plane circle of
radius r, there exists a unique angle z such that the area of the circle can be expressed by the
formula k2 tan2 z. Thus tan2z can be regarded as a scaling factor which measures the ratio of the
area of a circle to the area of the maximum triangle. But Bolyai observed that the only circles which
can be squared by compass and straightedge are those for which the number tan2z is either an
integer or a rational fraction whose denominator is of the form 2m p1 . p2 . p3 … , where m is a non-
negative integer and where each p is a (distinct) Fermat prime.

16. The Non-Euclidean Revolution

Tragically the revolutionary achievement of Lobachevsky and Bolyai in creating a new geometry was
largely met with indifference and neither received due recognition during his lifetime. This was due
to a number of factors:

1) the prestige enjoyed by Euclid whose geometry had gone unchallenged for over two
thousand years
2) the prevailing philosophical view, proposed by Immanuel Kant, that our (Euclidean)
perception of space is innate
3) the fact that Lobachevsky and Bolyai lived far removed from the great cultural centres of
Europe and their work was printed in obscure publications
 4) the difficulty, in the absence of a wider scheme of geometry, of seeing non-Euclidean
geometry as anything more than a curiosity.

Interest in non-Euclidean geometry only revived after Gauss’ death when his correspondence on the
subject, which contained references to the work of Lobachevsky and Bolyai, was published.
Mathematicians became interested in tracing the original papers, which within a few years were
translated into French and Italian.

14

Interest was further stimulated by the publication in 1868 of Bernhard Riemann’s ground-breaking
memoir On the Principles Underlying Geometry. Riemann (1822-1866) suggested that if Euclid’s
second postulate, that a straight line can be infinitely extended, is modified to state that the straight
line is finite but unbounded, while the parallel postulate is replaced by the assumption that no two
straight lines are parallel, then it is possible to create a geometry in which the sum of the angles of a
triangle is greater than two right angles, analogous to the geometry which holds true on the surface
of a sphere. The term elliptic was coined by Felix Klein to denote this type of non-Euclidean
geometry. Riemann’s memoir also provided a wider geometrical framework, which was previously
lacking, enabling both Euclidean and non-Euclidean geometry to be viewed in the context of what
are now called manifolds.

Then In 1868 Eugenio Beltrami (1835-1899) published his influential Essay on the Interpretation of
Non-Euclidean Geometry in which he showed that the hyperbolic plane can be partially mapped onto
the surface of a Euclidean surface of revolution, the pseudosphere. The pseudospehere , which was
first investigated in 1693 by Huygens, is a surface of negative curvature generated by rotating a
tractrix about its asymptote:

17. The Impossibility of Proving the Parallel Postulate

Gauss, Lobachevsky and Bolyai all suspected that it was not possible, except perhaps by some
physical experiment, either to establish or refute the validity of Euclidean geometry (see Appendix).
Henri Poincaré (1854-1912) finally demonstrated the logical consistency of non-Euclidean geometry
by mapping the hyperbolic plane onto an (open) Euclidean unit disc, in which straight lines are
represented either by diameters or the arcs of circles orthogonal to the circumference of the disc:

15

In order to map the whole of the hyperbolic plane onto the unit disc a suitable metric must be
introduced to ensure that arcs such as XAMP are infinitely long. This can be achieved by defining the
length of a segment such as AM by the metric

AM = log (AP.XM /XA.MP),

where AP, XA, MP, XM are chords of the arc XAMP. This definition of length ensures that a segment
AM becomes infinitely long as A approaches X or as M approaches P. In effect the circumference of
the disc is deemed to lie at infinity. The definition also satisfies the usual properties of a distance
function, among them the additive property, so that, given any point D (not shown) lying between A
and M, AM = AD + DM. This follows from the law of logarithms, whereby

AM = log (AP.XM/XA.MP) = log {(AP.XD/XA.DP).(DP.XM/XD.MP)} = log (J.K) = log J + log K = AD + DM

It is readily seen that all the postulates which are satisfied by straight lines in the hyperbolic plane
apply equally to orthogonal arcs on the Euclidean disc viz: 1) only one orthogonal arc can be drawn
through any two points on the disc 2) an arc can be extended indefinitely in either direction
3) a (non-orthogonal) circle of any radius can be drawn on the disc 4) all right angles on the disc are
equal and 5) the hyperbolic substitute for the parallel postulate applies to the orthogonal arcs. For
given an arc UV and a point A not on this arc, the arcs WAV and ZAU drawn through A represent the
left hand and right hand ‘parallels’ (so to speak) to UV, since each meets UV asymptotically at an
infinitely distant point at the circumference. In contrast other arcs drawn through A are either
intersecting with respect to UV, such as XAMP, or non-intersecting, such as SAT. Thus the ‘parallels’
WAV and ZAU separate the intersecting arcs from the non-intersecting arcs, just as boundary
parallels separate intersecting straight lines from non-intersecting straight lines in the hyperbolic
plane.

16

It follows that all the properties of hyperbolic geometry will apply to figures on the disc. For
example, the sum of the angles of any triangle on the disc formed by intersecting arcs, such as the
doubly asymptotic triangle AUV, is less than two right angles. Furthermore in the asymptotic right
angled triangle AMU, UAM (marked) can be considered as the angle of parallelism corresponding
to the segment AM. As a result it can be shown that the identity tan ½ UAM = e AM and all the
other trigonometric formulas of hyperbolic geometry hold true on the disc (where k is taken as equal
to 1). It is also possible to construct the equivalent of horocycles and equidistant lines on the disc.

The equivalence between the hyperbolic plane and the Euclidean disc shows that any contradiction
in hyperbolic geometry would also imply a contradiction in Euclidean geometry. It is therefore
impossible to establish the validity of one at the expense of the other. This in turn implies that
Euclid’s first four postulates cannot be used either to prove or disprove the parallel postulate. The
parallel postulate is independent of the other four, vindicating Euclid’s decision to place it among
the postulates, in contradiction to all those who wrongly treated it as an unproved theorem.

18. The Geometry of Space

For over 2000 years Euclid’s geometry had been regarded as synonymous with the geometry of
space and therefore the only conceivable geometry. (Spherical geometry was not considered an
alternative geometry since its theorems can be derived from Euclidean geometry). The discovery of
non-Euclidean geometry left geometry in the uncomfortable position of being a deductive science
based on assumptions which could not be validated either empirically or mathematically.

Yet paradoxically the very quandary in which geometry found itself released it from subordination to
the physical world. Mathematicians felt free thereafter to create new geometries, however abstract
they at first appeared to be. But once these geometries had been shown to be logically consistent,
they could be applied to the physical world an tested to see if they fitted the observed phenomena
An example is the way in which hyperbolic geometry has been used to model Einstein’s special
theory of relativity.
 17

Appendix

‘You must not attempt this approach to parallels. I know this way to its very end. I have traversed
this bottomless night, which extinguished all light and joy in my life. I entreat you, leave the science
of parallels alone.’

(Letter from Farkas Bolyai to his son Janos Bolyai, 1817)

‘I am resolved to publish a work on parallels as soon as I can put it in order, complete it, and the
opportunity arises. I have not yet made the discovery but the path which I have followed is almost
certain to lead to my goal, provided this goal is possible. I do not yet have it but I have found things
so magnificent that I was astounded. It would be an eternal pity if these things were lost as you, my
dear father, are bound to admit when you see them. All I can say now is that I have created a new
and different world out of nothing.’

(Letter from Janos Bolyai to Farkas Bolyai, 1823)

‘I have this day received from Hungary a little treatise on the Non-Euclidean Geometry wherein I
find again all my own ideas and results, developed with great elegance, howbeit for one to whom
the subject is strange, in a form somewhat difficult to follow because of the concentration. … I
regard this young geometer von Bolyai as a genius of the first order.’

(Letter from Gauss to Gerling, 14 February 1832)

‘If I commenced by saying that I am unable to praise this work, you would be surprised for a
moment. But I cannot say otherwise. To praise it would be to praise myself. Indeed the whole
contents of the work the path taken by your son, the results to which he is led, coincide almost
entirely with my meditations, which have occupied my mind partly for the last thirty or thirty-five
years. So I remain quite stupefied. So far as my own work is concerned, of which up till now I have
put very little on paper, my intention was not to let it be published in my lifetime. Indeed the
majority of people have not clear ideas upon the questions of which we are speaking, and I have
found very few people who could regard with any special interest what I communicated to them on
this subject. To be able to take such an interest it is first of all necessary to have devoted careful
thought to the real nature of what is wanted and upon this matter almost all are uncertain. On the
other hand it was my idea to write down all this later so that it should not perish with me. It is
therefore a pleasant surprise for me that I am spared this trouble, and I am very glad that it is just
the son of my old friend, who takes the precedence of me in such a remarkable manner.’

(Letter from Carl Gauss to Farkas Bolyai, 6 March 1832)

18

‘Gauss’ answer to your work is very satisfactory and redounds to the honour of our country and of
our nation.’

(Letter from Farkas Bolyai to Janos Bolyai, 1832)

‘In geometry I find certain imperfections which I hold to be the reason why this science, apart
from transitions into analytics, can as yet make no advance from that state in which it has come to
us from Euclid. As belonging to these imperfections, I consider … finally the momentous gap in the
theory of parallels, to fill which all efforts of mathematicians have so far been in vain.’

(Nikolai Lobachevsky, Geometrical Investigations, 1840)

‘Hence there is no means, other than astronomical observations, to use for judging of the
exactitude which pertains to the calculations of ordinary geometry’.

(Nikolai Lobachevsky, Geometrical Investigations, 1840)

‘I have found in Lobachevsky’s work nothing that is new to me, but the development is made in a
way different from that which I have followed, and certainly by Lobachevsky in a skilful way and in
truly geometric spirit. I feel that I must call your attention to the book, which will quite certainly
afford you the keenest pleasure.’

(Letter from Carl Gauss to Schumacher, 1846)

‘Even if in this remarkable work different methods are followed at times, nevertheless, the spirit
and result are much like the Appendix (The Absolute Science of Space) … If Gauss was, as he says,
surprised in the extreme, first by the Appendix and later by the striking agreement of the Hungarian
and Russian mathematicians, truly, none the less so am I.’
(Unpublished notes, Janos Bolyai, 1848)

19

Further Reading

*Seth Braver, Lobachevsky Illuminated, scholarworks.umt.edu (2007)

*V. Katz, A History of Mathematics, Wesley-Addison (2009) pp.839-851, UCL library online

Roberto Bonola, Non-Euclidean Geometry, Dover (1906, 1955)

H. E. Wolfe, Introduction to Non-Euclidean Geometry, Dryden (1945)

S. Kulczycki, Non-Euclidean Geometry, Pergamon (1981)

F. Kárteszi, Appendix Showing the Absolute Science of Space, North Holland Mathematical Studies
(1987)

Steven Rose, Notes on Bolyai’s Appendix, arxiv.org (2023),

Grateful acknowledge to H. Wolfe, S. Kulczycki and F. Kárteszi for the use of their texts in preparing
these notes.