A Spherical

Pythagorean
Theorem
PAOLO MARANER

here are probably many inequivalent statements

T
In right-angled triangles the square on the side opposite
in spherical geometry, somehow reducing to the to the right angle equals [the sum of] the squares on the
Pythagorean theorem in the limit of an infinite radius sides containing the right angle.
of curvature r. Among these, the Law of Cosines,
The words ‘the square on the side’ refer to the area of the
cosðc=rÞ ¼ cosða=rÞ cosðb=rÞ; square constructed on the side, which only incidentally
corresponds to ‘the square of the side’ in the sense of the
for a spherical right triangle with hypotenuse c and legs a and
second power of the length of the side. This correspondence
b, is generally presented as the ‘spherical Pythagorean the-
no longer holds in spherical or hyperbolic geometry, gen-
orem’. Still, it has to be remarked that this formula does not
erating not a little confusion about what the generalization of
have an immediate meaning in terms of areas of simple
the theorem should be. On the other hand, since in Euclidean
geometrical figures, as the Pythagorean theorem does. There
geometry the area of every regular polygon is proportional to
is no diagram that can be drawn on the surface of the sphere
the second power of the side, the change of preposition
to illustrate the statement in the spirit of ancient Greek
makes clear that the original Pythagorean squares can as well
geometry. In this note I reconsider the issue of extending the
be replaced by equilateral triangles, regular pentagons,
geometrical Pythagorean theorem to non-Euclidean geom-
regular hexagons or any other kind of regular polygon.
etries (with emphasis on the more intuitive spherical
Equivalently, since the area enclosed by the circle is again
geometry).1 In apparent contradiction with the statement
proportional to the second power of the diameter/radius, the
that the Pythagorean proposition is equivalent to Euclid’s
Pythagorean squares can also be replaced by circles with
parallel postulate, I show that such an extension not only
diameter/radius equal to the sides of the right triangle. The
exists, but also yields a deeper insight into the classical
reach of the Pythagorean theorem can be extended even
theorem.
further. In Proposition 31 of Book VI of the Elements, Euclid
The subject matter being familiar, I can dispense with
himself states that we are actually free to replace the squares
preliminaries and start right in with Euclid’s Elements [1].
with arbitrary shapes provided they are similar:

The Pythagorean Theorem In right-angled triangles the figure on the side opposite
The most celebrated theorem in mathematics [3] appears as to the right angle is equal to the similar and similarly
Proposition 47 of Book I of Euclid’s Elements. It says: described figures on the sides containing the right angle.

1
There is already a geometrical non-Euclidean generalization of the Pythagorean theorem [5], but it is not entirely satisfactory, because the figure on the hypotenuse is
made to depend on the figures on the sides.

46 THE MATHEMATICAL INTELLIGENCER Ó 2010 Springer Science+Business Media, LLC

 For regular polygons and circles, the choice is somehow
forced by symmetry. Not so for right triangles. The standard
and apparently natural choice of identifying the class of
plane right triangles with that of spherical right triangles
is unsatisfactory in many respects. In Euclidean geometry
the role of the right angle is unambiguous, and so is the
distinction between hypotenuse and legs. In spherical
geometry a triangle can have two or even three right
angles—and, correspondingly, two ‘hypotenuses’ and three
‘legs’ or three ‘hypotenuses’ and three ‘legs’. The very
... statement of the Pythagorean theorem makes little sense.
If one persists in treating right triangles, the existence in
spherical geometry of equilateral right triangles immediately
provides a counterexample to all Pythagorean statements:
The three figures constructed on the congruent sides are
Figure 1. Diagrams representing some of the infinitely many identical and the area of one of them can not equal the sum
equivalent variants of the Euclidean Pythagorean proposition. of the areas of the other two.
On the other hand, a plane right triangle can be char-
acterized in many different ways. Just to mention the most
We obtain infinitely many equivalent geometrical state-
obvious ones:
ments (see Figure 1), all summarized by the Pythagorean
formula c2 = a2 + b2, for any right triangle with hypote-
(a) a triangle with a right angle (whence the name);
nuse c and legs a and b.
(b) a triangle with an angle equal to a half of the sum of its
In spherical and hyperbolic geometry there is no con-
interior angles;
cept of similar figures. The areas of regular polygons with
(c) a triangle obtained by bisecting a rectangle (an equian-
equal sides are no longer proportional. Neither is the area of
gular quadrilateral, in preparation for non-Euclidean
the circle proportional to that of a regular polygon with side
geometries) by means of its diagonal;
equal to its diameter/radius or to that of another circle with
(d) an inscribed triangle having a diameter as a side.
radius equal to its diameter. All Pythagorean statements
become inequivalent and none of them remains associated
Each characterization potentially provides a different
with the Pythagorean formula. The question we pose is
generalization. The point is whether a generalization exists
whether at least one of these geometrical statements remains
satisfying at least one of the infinitely many Pythagorean
true when generalized to non-Euclidean geometries.
statements. To gain insight into this, let us briefly recon-
Clearly, any generalization based on similarity is meaning-
sider a few basic aspects of spherical geometry.
less, but what about the ones linked by symmetry? To answer
this question it is first necessary to decide what the gener-
alization of right triangles, regular polygons, and circles is.
Spherical Triangles
......................................................................... Spherical geometry can be obtained by replacing Euclid’s
fifth postulate with the statement that no parallel to a given
straight line can be drawn through a point not lying on it (in
PAOLO MARANER’S research is on differ-
AUTHOR

order to achieve a consistent system, however, the first and

ential geometry in physics. After a doctorate in
second postulates must also be partially modified). A model
Parma, he had postdoctoral appointments for such a geometry is the curved surface of a sphere of
there, at MIT and in Budapest. Since 2000, he arbitrary radius r: Straight lines are identified with great
has been a high-school teacher in Bolzano, also circles. On the sphere we can draw points, segments, angles,
teaching mathematics to economists at the triangles, every kind of polygon and circles. Spherical tri-
university. His side interests include running angles, in particular, come early on stage. They appear as
and swimming, and (as attested by the present Definition I of Book I of Menelaus’s2 Sphaerica [4]:
article) the history of mathematics.
A spherical triangle is the space included by arcs of great
School of Economics and Management circles on the surface of a sphere.
Free University of Bozen/Bolzano
via Sernesi 1, Bolzano, 39100 The absence of a strong notion of parallelism on the
Italy sphere invalidates a number of important results of
e-mail: [email protected] Euclidean geometry. Most remarkably, Proposition 32 of
Book I of Euclid’s Elements is replaced by:

2
Menelaus of Alexandria (c. 70–140 CE) was the first to use arcs of great circles instead of parallel circles on the sphere. This marked a turning point in the development
of spherical geometry. Being mainly interested in astronomical measurements and calculations, Menelaus did not consider theorems about area, like the Pythagorean
theorem.

Ó 2010 Springer Science+Business Media, LLC, Volume 32, Number 3, 2010 47

 In any spherical triangle the sum of the three interior C
angles is greater than two right angles.
Thus, in spherical geometry (a) above is not equivalent
to (b). This provides us with a first alternative generaliza-
A B
tion of plane right triangles to spherical geometry. O
The difference between the sum of the interior angles
and the straight angle
e ¼ sum of interior angles  p
is called the spherical excess of the triangle and is proved to Figure 3. Inscribed spherical triangles having a diameter as a
be proportional to the area A of the triangle itself, side are not right-angled.
A ¼ r2 e:
By triangulation these results straightforwardly extend to pþe
\ ABC ¼ :
every polygon: In any n-sided spherical polygon, the sum of 2
the n interior angles is greater than (2n - 4) right angles, and To prove the opposite implication, we just double a
the area of the polygon equals r2 times its spherical excess. In spherical triangle ABC with \ ABC ¼ pþe2 and join the two
particular, the sum of the four congruent interior angles of a copies along AC with A and C interchanged. Since \ BAC þ
spherical square is greater than four right angles. Hence, \ ACB ¼ pþe2 we obtain an equiangular quadrilateral.
these angles are no longer right. The triangulation of a To see that (d) implies (b), denote by e the spherical
spherical square by means of its diagonal no longer produces excess of the triangle ABC in Figure 3. Draw the segment
two right triangles. The same holds for every equiangular OC dividing ABC into two isosceles triangles AOC and
quadrilateral. It follows that (a) is not equivalent to (c). This BOC. Denote by e1 the spherical excess of the first one and
provides us with a second possible generalization of plane by e2 that of the second one. Clearly, e = e1 + e2. Since
right triangles to spherical geometry. \OCA  \OAC, from the first triangle, we obtain
A third possibility comes from the failure of Proposition 2\OCA þ \ AOC ¼ p þ e1 ;
20 of Book III of Euclid’s Elements and of its corollaries. In
particular: and since \OCB  \OBC, from the second one we have
2\OCB þ \ BOC ¼ p þ e2 :
In a given spherical circle, all inscribed angles sub-
tending the diameter are greater than a right angle. Adding term by term, recalling that \ ACO þ \ BCO 
\ ACB and \ AOC þ \ BOC ¼ p, we obtain
Inscribed angles subtending the diameter are no longer pþe
right. Therefore, in spherical geometry (a) is not equivalent \ ACB ¼ :
2
to (d).
Quite remarkably, in spherical geometry (b), (c), and (d) Finally, to prove that (b) implies (d), we consider a
are equivalent. spherical triangle ABC with \ ACB ¼ pþe 2 . We now choose
To see that (c) implies (b), consider Figure 2. Since point O on AB such that \ ACO equal to \ BAC. Thus,
equiangular quadrilaterals have opposite sides congruent, CO:AO. At this point, we observe that \ BCO ¼
ABC and ACD are congruent. Denote by e their spherical pþe pþe
2  \ ACO ¼ 2  \ BAC ¼ \CBA. Thus, CO:BO, and
excess. Since spherical excess is proportional to the area the triangle ABC is inscribed in a circle with diameter AB.
and the area of the equilateral quadrilateral ABCD equals The transition from Euclidean to spherical geometry
the sum of the areas of the triangles ABC and ACD, the seems to preserve the property of ‘having one angle equal
spherical excess of the equiangular quadrilateral equals 2e. to a half of the sum of its interior angles’ and not the
The sum of its interior angles is therefore 2p + 2e. Given property of ‘having a right angle’. This provides us with a
the congruence of the four interior angles, we obtain promising class of triangles generalizing plane right trian-
gles to non-Euclidean geometries. Let us therefore
introduce a suitable terminology:

D C We say that a triangle is properly angled, or, equiva-

lently, that it is a proper triangle, when it has an angle
equal to a half of the sum of its interior angles. That
angle is called the proper angle of the triangle; the side
opposite to it, the hypotenuse; and the sides containing it
A B the legs.

Figure 2. Spherical triangles obtained by dividing an equian- The role of the proper angle is unambiguous, and so is
gular quadrilateral by means of its diagonal are not right- the distinction between hypotenuse and legs. In plane
angled. geometry the class of proper triangles corresponds to that

48 THE MATHEMATICAL INTELLIGENCER

of right triangles. In spherical geometry the class of proper we evaluate the lengths of the sides as
triangles shares at least some of the fundamental properties
enjoyed by plane right triangles: Any equiangular quadri- AB ¼ r cos1 ðcos2 ^
h þ sin2 ^ ^
h cos /Þ;
lateral is divided by means of its diagonal into two proper BC ¼ r cos1 ðcos2 ^
h  sin2 ^ ^
h cos /Þ;
triangles; an inscribed triangle having as side a diameter is a
proper triangle. It is then natural to wonder whether AC ¼ r cos1 ðcos2 ^
h  sin2 ^
hÞ:
spherical proper triangles enjoy at least one of the infinitely
Dividing by r and taking the cosine of the resulting
many symmetric variants of the Pythagorean proposition.
expressions we have
Recalling the formula expressing the area of a spherical
regular polygon of side l,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðAB=rÞ ¼ cos2 ^
h þ sin2 ^ ^
h cos /;
2 2 1 cosðl=rÞ  cosð2p=nÞ
Angon ¼ 2pr  2nr sin cosðBC=rÞ ¼ cos2 ^
h  sin2 ^ ^
h cos /;
cosðl=rÞ þ 1
cosðAC=rÞ ¼ cos2 ^
h  sin2 ^
h:
and the formula for the area of a spherical circle of radius r
Adding the first two equalities and comparing the result
Acircle ¼ 2pr2 ð1  cosðr=rÞÞ;
with the third one, after a very little algebra we obtain
we can simply proceed to a direct check of all of them. It is a
2pr2 ð1  cosðAC=rÞÞ ¼ 2pr2 ð1  cosðAB=rÞÞ
wonderful surprise to discover that one of them still holds
true. þ 2pr2 ð1  cosðBC=rÞÞ:

Recalling the formula for the area of the spherical circle in

Pythagoras on the Sphere ...
terms of its radius, we recognize the spherical Pythagorean
To pay homage to ancient Greek geometers, we state the
proposition. Clearly, in the limit of a large radius of curvature
proposition as follows:
r, this expression reduces to the Pythagorean formula
2 2 2
AC ¼ AB þ BC :
In properly angled triangles, the circle on the side
opposite to the proper angle equals [the sum of] the
circles on the sides containing the proper angle. ... and on the Hyperbolic Plane
The proposition straightforwardly extends to the less intuitive
Here, the words ‘the circle on the side’ mean the area of hyperbolic geometry. This is proved pretty much in the same
the circle having the side as radius; this time there is no risk way. As hyperbolic plane model we consider the quadric
of algebraic confusion.
x 2 þ y 2  z 2 ¼ r2
The proposition is illustrated by the beautiful diagram of
Figure 4. It is also immediate how to give an analytical proof embedded in the Minkowskian space R2;1 . By introducing
of it. Parametrizing the sphere by standard spherical coor- hyperbolic polar coordinates
dinates h and /, we consider an arbitrary equiangular !
quadrilateral ABCD centered at the north pole and with x ¼ ðr sinh w cos /; r sinh w sin /; r cosh wÞ;
diagonal on the great circle through the pole and (1, 0). Its the plane is parametrized by the hyperbolic latitude w,
vertices lie at Að^h; 0Þ, Bð^ ^ Cð^
h; /Þ, h; pÞ, Dð^ ^  pÞ, for some
h; / w C 0, and by the longitude /, - p \ / B p. The distance
angles ^h and /.^ Given the equivalence of (b) and (c), ABC is
formula for generic points P(wP, /P), Q(wQ, /Q) reads
an arbitrary proper triangle. By means of the spherical dis-
tance formula for generic points PðhP ; /P Þ and Q(hQ, /Q), PQ ¼ r cosh1
   
PQ ¼ r cos1 cos hP cos hQ þ sin hP sin hQ cosð/Q  /P Þ ; cosh wP cosh wQ  sinh wP sinh wQ cosð/Q  /P Þ :

As in spherical geometry, proper triangles are obtained by

dividing equiangular quadrilaterals by means of their diag-
onals. Hence, we again consider an arbitrary equiangular
quadrilateral ABCD centered at the pole (0, 0), with diagonal
along the hyperbolic line through the pole and (1, 0). The
vertices lie at Aðw;^ 0Þ, Bðw; ^ Cðw;
^ /Þ, ^ pÞ, Dðw;^ /^  pÞ, for
^ ^
some values w and / . ABC is an arbitrary proper triangle. The
lengths of its sides are evaluated as

AB ¼ r cosh1 ðcosh2 w
^  sinh2 w ^
^ cos /Þ;
BC ¼ r cosh1 ðcosh2 w
^ þ sinh2 w ^
^ cos /Þ;
AC ¼ r cosh1 ðcosh2 w
^ þ sinh2 wÞ:
^

Dividing by r, taking the hyperbolic cosine of the

three expressions, and recalling the identity sinh2 x ¼
Figure 4. The spherical Pythagorean proposition. cosh2 x  1, after some algebra we obtain

Ó 2010 Springer Science+Business Media, LLC, Volume 32, Number 3, 2010 49

 2pr2 ðcoshðAC=rÞ  1Þ ¼ 2pr2 ðcoshðAB=rÞ  1Þ first four Euclidean postulates. Since it is also true in
þ 2pr2 ðcoshðBC=rÞ  1Þ: spherical geometry, the statement should actually follow
from an even smaller set of axioms. In any case, it repre-
Recalling the formula for the area of an hyperbolic circle sents a more basic theorem about area than the original
of radius r Pythagorean theorem (as in Euclidean geometry, spherical
and hyperbolic polygons of the same area are related by
Acircle ¼ 2pr2 ðcoshðr=rÞ  1Þ;
scissor congruence [2]).
we recognize the hyperbolic Pythagorean proposition. The In this paper we presented an analytical proof of the
Euclidean Pythagorean formula is again obtained in the spherical and hyperbolic Pythagorean propositions. In the
limit of a large radius of curvature r. final analysis, this proof follows from the Euclidean
Pythagorean proposition itself. It goes without saying that a
synthetic proof based on a minimal choice of postulates
Epilogue would be of great interest.
The Pythagorean theorem is generally claimed to be
equivalent to Euclid’s fifth postulate. If so, then it can hold
only in Euclidean geometry. As we have seen in this paper, REFERENCES
this very much depends on how the proposition is under- [1] Euclid, The Elements, translated with introduction and commen-
stood. If we insist on squares on the sides of right triangles, tary by T. L. Heath, Dover, New York, 1956.
no doubt the claim is true. Nevertheless, if we take a [2] R. Hartshorne, Geometry: Euclid and Beyond, Springer-Verlag,
slightly wider viewpoint by considering all the equivalent New York, 2000.
variants of the theorem, and classes of triangles that better [3] E. Maor, The Pythagorean Theorem, Princeton, Princeton and
embody the properties of plane right triangles in non- Oxford, 2007.
Euclidean geometry, we come to a statement that equally [4] Menelaus, Sphaerica, translated into Latin from the Arabic version
holds in Euclidean, spherical, and hyperbolic geometry. by E. Halley, http://www.wilbourhall.org
Since it is true in Euclidean and hyperbolic geometry, [5] C. Piel, Der Lehrsatz des Pythagoras in der hyperbolischen
this statement belongs to neutral geometry. In principle, it Geometrie, Arch. Math. Phys. (1914) 22 199–204.
could be included among the first 28 propositions of the
Elements and should be capable of a proof in terms of the

50 THE MATHEMATICAL INTELLIGENCER

Copyright of Mathematical Intelligencer is the property of Springer Science & Business Media B.V. and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.