Papers on Topology Analysis Situs and Its Five

Supplements Henri Poincare pdf download

https://ebookgate.com/product/papers-on-topology-analysis-situs-
and-its-five-supplements-henri-poincare/

Get Instant Ebook Downloads – Browse at https://ebookgate.com

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Henri Poincare A Biography through the Daily Papers 1st

Edition Jean-Marc Ginoux

https://ebookgate.com/product/henri-poincare-a-biography-through-the-
daily-papers-1st-edition-jean-marc-ginoux/

ebookgate.com

Selected Papers on Ancient Literature and its Reception 2

Bände mit 1 ISBN Edition Philip Hardie

https://ebookgate.com/product/selected-papers-on-ancient-literature-
and-its-reception-2-bande-mit-1-isbn-edition-philip-hardie/

ebookgate.com

Brookings Papers on Economic Activity 2005 Brookings

Papers on Economic Activity William C. Brainard

https://ebookgate.com/product/brookings-papers-on-economic-
activity-2005-brookings-papers-on-economic-activity-william-c-
brainard/
ebookgate.com

Winding Around The Winding Number in Topology Geometry and

Analysis 1st Edition John Roe

https://ebookgate.com/product/winding-around-the-winding-number-in-
topology-geometry-and-analysis-1st-edition-john-roe/

ebookgate.com
Handbook of Petroleum Product Analysis Chemical Analysis A
Series of Monographs on Analytical Chemistry and Its
Applications 1st Edition James G. Speight
https://ebookgate.com/product/handbook-of-petroleum-product-analysis-
chemical-analysis-a-series-of-monographs-on-analytical-chemistry-and-
its-applications-1st-edition-james-g-speight/
ebookgate.com

Five eyes on the fence protecting the five core capitals

of your business First Edition Rose

https://ebookgate.com/product/five-eyes-on-the-fence-protecting-the-
five-core-capitals-of-your-business-first-edition-rose/

ebookgate.com

Nutrients dietary supplements and nutriceuticals Cost

analysis versus clinical benefits 1st Edition Jørgen
Dejgaard Jensen (Auth.)
https://ebookgate.com/product/nutrients-dietary-supplements-and-
nutriceuticals-cost-analysis-versus-clinical-benefits-1st-edition-
jorgen-dejgaard-jensen-auth/
ebookgate.com

Quantum spaces Poincare Seminar 2007 1st Edition Vincent

Rivasseau

https://ebookgate.com/product/quantum-spaces-poincare-
seminar-2007-1st-edition-vincent-rivasseau/

ebookgate.com

Non Hausdorff Topology and Domain Theory Selected Topics

in Point Set Topology 1st Edition Jean Goubault-Larrecq

https://ebookgate.com/product/non-hausdorff-topology-and-domain-
theory-selected-topics-in-point-set-topology-1st-edition-jean-
goubault-larrecq/
ebookgate.com
 Papers on Topology
Analysis Situs and Its Five Supplements

Henri Poincaré
Translated by John Stillwell

July 31, 2009

2
Contents

Translator’s Introduction 1
Topology before Poincaré . . . . . . . . . . . . . . . . . . . . . . . . . 1
Poincaré before topology . . . . . . . . . . . . . . . . . . . . . . . . . . 3
The Analysis situs paper . . . . . . . . . . . . . . . . . . . . . . . . . 5
The five supplements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
The Poincaré conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Comments on terminology and notation . . . . . . . . . . . . . . . . . 11
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

On Analysis Situs 15

Analysis Situs 18
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
§1. First definition of manifold . . . . . . . . . . . . . . . . . . . . . . 20
§2. Homeomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
§3. Second definition of manifold . . . . . . . . . . . . . . . . . . . . . 24
§4. Oppositely oriented manifolds . . . . . . . . . . . . . . . . . . . . . 28
§5. Homologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
§6. Betti numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
§7. The use of integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 32
§8. Orientable and non-orientable manifolds . . . . . . . . . . . . . . . 34
§9. Intersection of two manifolds . . . . . . . . . . . . . . . . . . . . . 40
§10. Geometric representation . . . . . . . . . . . . . . . . . . . . . . . 49
§11. Representation by a discontinuous group . . . . . . . . . . . . . . 54
§12. The fundamental group . . . . . . . . . . . . . . . . . . . . . . . 58
§13. Fundamental equivalences . . . . . . . . . . . . . . . . . . . . . . 60
§14. Conditions for homeomorphism . . . . . . . . . . . . . . . . . . . 65
§15. Other modes of generation . . . . . . . . . . . . . . . . . . . . . . 75
§16. The theorem of Euler . . . . . . . . . . . . . . . . . . . . . . . . . 85
§17. The case where p is odd . . . . . . . . . . . . . . . . . . . . . . . 93
§18. Second proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

3
4 Contents

Supplement to Analysis Situs 100

§I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
§II. Schema of a polyhedron . . . . . . . . . . . . . . . . . . . . . . . . 103
§III. Reduced Betti numbers . . . . . . . . . . . . . . . . . . . . . . . 107
§IV. Subdivision of polyhedra . . . . . . . . . . . . . . . . . . . . . . . 111
§V. Influence of subdivision on reduced Betti numbers . . . . . . . . . 112
§VI. Return to the proofs of paragraph III . . . . . . . . . . . . . . . . 117
§VII. Reciprocal polyhedra . . . . . . . . . . . . . . . . . . . . . . . . 120
§VIII. Proof of the fundamental theorem . . . . . . . . . . . . . . . . . 126
§IX. Various remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
§X. Arithmetic proof of a theorem of paragraph VII . . . . . . . . . . 133
§XI. The possibility of subdivision . . . . . . . . . . . . . . . . . . . . 136

Second Supplement to Analysis Situs 141

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
§1. Review of the principal definitions . . . . . . . . . . . . . . . . . . 141
§2. Reduction of tables . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
§3. Comparison of the tables Tq and Tq0 . . . . . . . . . . . . . . . . . 149
§4. Application to some examples . . . . . . . . . . . . . . . . . . . . . 152
§5. Extension to the general case of a theorem in the first supplement 158
§6. Internal torsion of manifolds . . . . . . . . . . . . . . . . . . . . . 165

Third Supplement to Analysis Situs 170

Fourth Supplement to Analysis Situs 187

§1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
§2. Three-dimensional cycles . . . . . . . . . . . . . . . . . . . . . . . 191
§3. Two-dimensional cycles . . . . . . . . . . . . . . . . . . . . . . . . 200
§4. One-dimensional cycles . . . . . . . . . . . . . . . . . . . . . . . . 209
§5. Various remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

Fifth Supplement to Analysis Situs 220

§1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
§2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
§3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
§4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
§5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
§6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Translator’s Introduction

Topology before Poincaré

Without much exaggeration, it can be said that only one important topological
concept came to light before Poincaré. This was the Euler characteristic of
surfaces, whose name stems from the paper of Euler (1752) on what we now call
the Euler polyhedron formula. When writing in English, one usually expresses
the formula as
V − E + F = 2,
where
V = number of vertices,
E = number of edges,
F = number of faces,
of a convex polyhedron or, more generally, of a subdivided surface homeomor-
phic to the two-dimensional sphere S2 . In Poincaré (for example, in §16 of his
Analysis situs paper) one finds the French version
S − A + F = 2,
where S stands sommets and A for arêtes.
The formula is usually proved by showing that the quantity V −E+F remains
invariant under all possible changes from one subdivision to another. It follows
that V − E + F is an invariant of any surface, not necessarily homeomorphic
to S2 . This invariant is now called the Euler characteristic. Thus S2 has Euler
characteristic 2, whereas the torus has Euler characteristic 0.
Between the 1820s and 1880s, several different lines of research were found
to converge to the Euler characteristic.
1. The classification of polyhedra, following Euler (and even before him,
Descartes). Here the terms “edges” and “faces” have their traditional
meaning in Euclidean geometry.
2. The classification of surfaces of constant curvature, where the “edges” are
now geodesic segments. Here one finds that surfaces of positive Euler char-
acteristic have positive curvature, those of zero Euler characteristic have

1
2 Translator’s Introduction

zero curvature, and those of negative Euler characteristic have negative

curvature.

3. More generally, the average curvature of a smooth surface is positive for

positive Euler characteristic, zero for zero Euler characteristic, and nega-
tive for negative Euler characteristic. This follows from the Gauss-Bonnet
theorem of Gauss (1827) and Bonnet (1848).

4. The study of algebraic curves, revolutionized by Riemann (1851) when

he modelled each complex algebraic curve by a surface—its “Riemann
surface.” Under this interpretation, a number that Abel (1841) called the
genus of an algebraic curve turns out to depend on the Euler characteristic
of its Riemann surface. In fact, genus g is related to Euler characteristic
χ by
χ = 2 − 2g.
Moreover, the genus g has a simple geometric interpretation as the number
of “holes” in the surface. Thus S2 has genus 0 and the torus has genus 1.

5. The topological classification of surfaces, by Möbius (1863). Möbius stud-

ied closed surfaces in R3 by slicing them into simple pieces by parallel
planes. He found by this method that every such surface is homeomor-
phic to a standard surface with g holes. Thus closed surfaces in R3 —that
is, all orientable surfaces—are classified by their genus, and hence by their
Euler characteristic. (Despite his discovery of the non-orientable surface
that bears his name, Möbius did not classify non-orientable surfaces. This
was done by Dyck (1888).)

6. The study of “pits, peaks, and passes” on surfaces in R3 by Cayley (1859)

and Maxwell (1870). A family of parallel planes in R3 intersects a surface
S in curves we may view as curves of “constant height” (contour lines)
on S. If the planes are taken to be in general position, and the surface is
smooth, then S has only finitely many “pits, peaks, and passes” relative
to the height function. It turns out that

number of peaks − number of passes + number of pits

is precisely the Euler characteristic of S.

All of these ideas admit generalizations to higher dimensions, but the only
substantial step towards topology in arbitrary dimensions before Poincaré was
that of Betti (1871). Betti was inspired by Riemann’s concept of connectivity of
surfaces to define connectivity numbers, now known as Betti numbers P1 , P2 , . . .,
in all dimensions. The connectivity number of a surface S may be defined as
the maximum number of disjoint closed curves that can be drawn on S without
separating it. This number P1 is equal to the genus of S, hence it is just the
Euler characteristic in disguise.
Poincaré before topology 3

For a three-dimensional manifold M one can also consider the maximum

number P2 of disjoint closed surfaces in M that fail to separate M as the “two-
dimensional connectivity number” of M . The idea of separation fails to explain
the “one-dimensional connectivity” of M , however, since no finite set of curves
can separate M . Instead, one takes the maximum number of curves that can
lie in M without forming the boundary of a surface in M . (For a surface M ,
this maximum is the same as Riemann’s connectivity number.) Betti defined
Pm similarly, in a manifold M of arbitrary dimension, as the maximum number
of m-dimensional pieces of M that do not form the boundary of a connected
(m + 1)-dimensional piece of M . Thus Betti brought the concept of boundary
into topology in order to generalize Riemann’s concept of connectivity.
This was Poincaré’s starting point, but he went much further, as we will see.

Poincaré before topology

In the introduction to his first major topology paper, the Analysis situs, Poincaré
(1895) announced his goal of creating of creating an n-dimensional geometry.
As he memorably put it:

. . . geometry is the art of reasoning well from badly drawn figures;

however, these figures, if they are not to deceive us, must satisfy
certain conditions; the proportions may be grossly altered, but the
relative positions of the different parts must not be upset.

Because “positions must not be upset,” Poincaré sought what Leibniz called
Analysis situs, a geometry of position, or what we now call topology. He cited
as precedents the work of Riemann and Betti, and his own experience with
differential equations, celestial mechanics, and discontinuous groups. Of these,
I believe the most influential was the last, which stems from his work (and the
related work of Klein) on fuchsian functions in early 1880s.
Poincaré’s major papers on fuchsian functions may be found translated into
English in Poincaré (1985). The ideas relevant to topology may be summarized
as follows.
One considers a group Γ of translations of the plane which is fixed-point-free
(that is, non-identity group elements move every point) and discontinuous (that
is, there is a non-zero lower bound to the distance that each point is moved by
the non-identity elements). A special case is where the plane is C and Γ is
generated by two Euclidean translations in different directions. Generally the
“plane” is the hyperbolic plane H2 , which may be modeled by either the upper
half plane of C or the open unit disk {z : |z| < 1}.
In either case, Γ has a fundamental domain D which is a polygon, and
the plane is filled without overlapping by its translates γD for γ ∈ Γ. In the
special case where the plane is C the fundamental domain can be taken to be
a parallelogram, the translations of which in the two directions fill C. In the
hyperbolic case D is a polygon with 4g sides for some g ≥ 2, and Γ is generated
4 Translator’s Introduction

by 2g elements, each of which translates D to a polygon with just one side in

common with D.
It follows that the quotient C/Γ in the special case is a torus (obtained
by identifying opposite sides of the fundamental parallelogram), while in the
hyperbolic case the quotient H2 /Γ is a surface of genus g ≥ 2, obtained by
identifying sides of the fundamental 4g-gon in certain pairs.1
Each pair of identified sides come together on the quotient surface as a closed
curve. For example, the identified sides of a fundamental parallelogram become
two closed curves a and b on the torus C/Γ, as shown in Figure 1. The curve a

a a → → a
b

Figure 1: Constructing a torus from a fundamental parallelogram.

corresponds to the translation of C with direction and length of the sides marked
a of the parallelogram, and curve b similarly corresponds to the translation with
direction and length of the sides marked b. Thus one is led to think of a “group
of curves” on the torus, isomorphic to the group of translations of C generated
by the translations a and b.
This group is what Poincaré later called the fundamental group, and we can
see why he viewed it as a group of “substitutions”—in this case, translations
of C—rather than as a group of (homotopy classes of) closed curves with fixed
origin on the torus, as we now do. Indeed, Poincaré in the 1880s much preferred
to work with fundamental polygons in the plane, and it was Klein (1882) who
realized that insight could be gained by looking at the quotient surface instead.
In particular, Klein used the Möbius classification of surfaces into canonical
forms to find canonical defining relations for fuchsian groups.
Here is how we find the defining relation in the case of the torus (in which
case one relation suffices). Clearly, if we perform the translations a, b, a−1 , b−1 of
C in succession, where a−1 and b−1 denote the inverses of a and b respectively,
the whole plane arrives back at its starting position. We write this relation
symbolically as
aba−1 b−1 = 1,

where 1 denotes the identity translation. On the torus surface, aba−1 b−1 denotes
a closed curve that bounds a parallelogram (reversing the process shown in
Figure 1), and hence this curve is contractible to a point. This is the topological
1 The connection with fuchsian functions, often mentioned by Poincaré but not very relevant

to topology, is that there are functions f that are periodic with respect to substitutions from
the group Γ: that is, f (γ(z)) = f (z) for all γ ∈ Γ. In the special case of the torus these
functions are the famous elliptic functions.
The Analysis situs paper 5

interpretation of the relation

aba−1 b−1 = 1.

With either interpretation it is quite easy to show that all relations between
a and b follow from the single relation aba−1 b−1 = 1. This is why we call
aba−1 b−1 = 1 the defining relation of the torus group.
In a similar way, we find the defining relation of any surface group by cutting
the surface along closed curves so as to produce a polygon. Equating the se-
quence of edges in the boundary of this polygon to 1 then gives a valid relation,
and again it is not hard to show that the relation thus obtained is a defining
relation. The curves most commonly used for the surface Sg of genus g are
called a1 , b1 , a2 , b2 , . . . , ag , bg , and cutting Sg along them produces a polygon
with boundary
a1 b1 a−1 −1 −1 −1
1 b1 · · · ag bg ag bg .

Consequently, the group of Sg may be generated by elements a1 , b1 , . . . , ag , bg

and it has defining relation

a1 b1 a−1 −1 −1 −1
1 b1 · · · ag bg ag bg = 1.

Figure 2 shows the curves a1 , b1 , a2 , b2 on the surface S2 , and the resulting

polygon.

a1
b1 b1

b2 a1

−→ a1 a2

a2 b1

b2 b2
a2

Figure 2: Genus 2 surface and its fundamental polygon

The Analysis situs paper

Poincaré set the agenda for his 1895 Analysis situs paper with a short announce-
ment, Poincaré (1892), a translation of which is also included in this volume.
In it he raises the question whether the Betti numbers suffice to determine the
topological type of a manifold, and introduces the fundamental group to further
illuminate this question. He gives a family of three-dimensional manifolds, ob-
tained as quotients of R3 by certain groups with a cube as fundamental region,
6 Translator’s Introduction

and shows that certain of these manifolds have the same Betti numbers but
different fundamental groups. It follows, assuming that the fundamental group
is a topological invariant, that the Betti numbers do not suffice to distinguish
three-dimensional manifolds.
In Analysis situs, Poincaré develops these ideas in several directions.
1. He attempts to provide a new foundation for the Betti numbers in a rudi-
mentary homology theory, which introduces the idea of computing with
topological objects (in particular, adding, subtracting, testing for linear
independence). As Scholz (1980), p. 300, puts it:
The first phase of algebraic topology, inaugurated by Poincaré,
is characterized by the fact that its algebraic relations and op-
erations always deal with topological objects (submanifolds).
2. Using his homology theory, he discovers a duality theorem for the Betti
numbers of an n-dimensional manifold:
Pm = Pn−m for m = 1, 2, . . . , n − 1.
In words: “ the Betti numbers equidistant from the ends are equal.” He
later called this the fundamental theorem for Betti numbers (p. 125).
3. He generalizes the Euler polyhedron formula to arbitrary dimensions and
situates it in his homology theory.
4. He constructs several three-dimensional manifolds by identifying faces of
polyhedra, observing that this leads natural presentations of their funda-
mental groups by generators and relations.
5. Recognizing that the fundamental group first becomes important for three-
dimensional manifolds, Poincaré asks whether it suffices to distinguish
between them. He is not able to answer this question.
Analysis situs is rightly regarded as the origin of algebraic topology, because
of Poincaré’s construction of homology theory and the fundamental group. The
fundamental group is the more striking of the two, because it is a blatantly
abstract structure and generally non-commutative, yet surprising easy to grasp
via generators and relations. Homology theory reveals an algebraic structure
behind the bare Betti numbers and Euler characteristics of Poincaré’s predeces-
sors, but it is not easy to say what this structure really is. Indeed, Poincaré
did not realize that the Betti numbers are only part of the story, and he had to
write Supplements 1 and 2 to Analysis situs before the so-called torsion coeffi-
cients came to light. And it was only in 1925 that Emmy Noether discovered
the homology groups, which we now view as the proper home of the Betti and
torsion numbers. She announced this discovery in Noether (1926).2
2 To be fair to Poincaré, he came close to discovering the first homology group H as
1
the abelianisation of the fundamental group π1 in Analysis situs §13. There he considered
the homologies obtained by allowing the generators of π1 to commute, and observed that
“knowledge of these homologies immediately yields the Betti number P1 .”
The five supplements 7

Thus, along with great breakthroughs, there is also confusion in Analysis

situs. The confusion extends to the very subject matter of algebraic topology,
the manifolds (or “varieties” as Poincaré calls them). His definitions suggest
that he is generally thinking of differentiable manifolds, but most of his three-
dimensional examples are defined combinatorially, by identifying faces of poly-
hedra, without checking their differentiability. His definition of Betti numbers
needs revision, as he discovers in Supplements 1 and 2, and imprecise arguments
are frequently used.
Another source of confusion concerns “simply-connected manifolds,” and it
ultimately led to the famous Poincaré conjecture in Supplement 5. In Analysis
situs, §14, he defines a manifold to be simply connected if its fundamental group
is trivial. It follows easily that a sphere of any dimension is simply-connected,
but Poincaré sometimes forgets that the converse is not obvious. On occasion, he
assumes that any simply-connected manifold is homeomorphic to a sphere (for
example, on p. 141), and on other occasions he even assumes (wrongly) that a
region with trivial homology is simply-connected (for example, on p. 59). These
errors have been flagged by footnotes, some by the original editors of Volume
VI of Poincaré’s Œuvres, René Garnier and Jean Leray (the actual author of
these footnotes is not identified), and some by myself.
Actually, it is not surprising that Poincaré made mistakes, given the novelty
and subtle nature of the subject, and his style of work. Darboux (1952), p. lvi,
describes Poincaré’s working method as follows:

Whenever asked to resolve a difficulty, his response came with the

speed of an arrow. When he wrote a memoir, he drafted it all in
one go, with only a few erasures, and did not return to what he had
written.

It is perhaps wise to read Poincaré’s memoirs in the same style: try to take in
their general sweep without lingering too long over gaps and errors.

The five supplements

In 1899 Poincaré wrote Complément a l’analysis situs in response to the criti-
cism of Heegaard (1898). Heegaard had become interested in three-dimensional
manifolds, and he found an example where Poincaré’s definition of Betti num-
bers comes into conflict with his duality theorem. To save the theorem, Poincaré
revised his homology theory in the Complément, moving towards a more combi-
natorial theory in which manifolds are assumed to have a polyhedral structure,
and computing Betti numbers from the incidence matrices of this structure. He
also arrived at a clearer explanation of the duality theorem in terms of the dual
(“reciprocal”) subdivision of a polyhedron, in which cells of dimension m in
the original polyhedron correspond to cells of dimension n − m in its dual. He
concluded the Complement with a (rather unconvincing) attempt to prove that
every differentiable manifold has a polyhedral subdivision. This theorem was
first proved rigorously by Cairns (1934).
8 Translator’s Introduction

Poincaré may have thought that the Complément would complete his Anal-
ysis situs paper, but four more “complements” were to follow, as further gaps
and loose ends came to light. For this reason, I have chosen to use the word
“supplement” rather than “complement” (as Poincaré himself did on occasion).
In the second supplement, Poincaré dug more deeply into the problems of his
original homology theory, uncovering the existence of torsion, and expanding his
technique for computing Betti numbers to one that also computes torsion coef-
ficients. He motivated his choice of the word “torsion” by showing that torsion
occurs only in manifolds, such as the Möbius band, that are non-orientable and
hence “twisted onto themselves” in some way (p. 168). When Emmy Noether
built the Betti numbers and torsion numbers into the homology groups in 1926,
the word “torsion” took up residence in algebra, much to the mystification of
group theory students who were not informed of its origin in topology.
Having now attained some mastery of homology theory, Poincaré was em-
boldened to conjecture (p. 169) that: the three-dimensional sphere is the only
closed three-dimensional manifold with trivial Betti and torsion numbers. This
was his first (and incorrect) version of the Poincaré conjecture.
The third and fourth supplements hark back to the first major application
of Betti numbers to classical mathematics, the work of Picard (1889) on the
connectivity of algebraic surfaces. An algebraic surface (or “algebraic function
of two variables” as Picard called it) is taken to have complex values of the
variables, hence it has four real dimensions. By a mixture of analytic and
topological arguments, Picard succeeded in finding the first Betti number P1
of algebraic surfaces, but he had less success in finding P2 . Invoking his new
homology theory, Poincaré pushed on to P2 in his fourth supplement (as far he
needed to go, since P3 = P1 , by Poincaré duality). Like Picard, Poincaré also
appealed to results from analysis, in his case referring to his work on fuchsian
functions and non-euclidean geometry from the early 1880s. An exposition of
Poincaré’s argument (in German) may be found in Scholz (1980), pp. 365–371.
The return to non-euclidean geometry paid off unexpectedly in the fifth
supplement, with an interesting geometric algorithm (p. 245) to decide whether
a curve on a surface is homotopic to a simple curve. Poincaré’s result is that,
in the case of genus greater than 1 where the surface can be given a non-
euclidean metric, a homotopy class contains a simple curve if and only the
geodesic representative is simple. Informally speaking, one can decide whether
a curve κ is homotopic to simple curve by “stretching κ tight” on the surface
and observing whether the stretched form of κ is simple. It seems likely that
Poincaré’s application of non-euclidean geometry to surface topology inspired
the later work of Dehn and Nielsen between 1910 and the 1940s, and the work
of Thurston in the 1970s.
In the fifth supplement, the result on simple curves is just part of a rather
meandering investigation of curves on surfaces, and their role in the construction
of three-dimensional manifolds (“Heegaard diagrams”). In the final pages of the
paper this investigation leads to a spectacular discovery: the Poincaré homology
sphere. By pasting together two handlebodies of genus 2, H1 and H2 say, so
that certain carefully chosen curves on H1 become identified with canonical disk-
The Poincaré conjecture 9

spanning curves on H2 , Poincaré obtains a three-dimensional manifold V whose

fundamental group π1 (V ) he can write down in terms of generators and relations.
To the reader’s astonishment, the presentation of π1 (V ) implies relations that
hold in the icosahedral group, so π1 (V ) is non-trivial. On the other hand,
by allowing the generators of π1 (V ) to commute one finds (in our language)
that H1 (V ) = 0, so that V is a closed three-dimensional manifold with trivial
homology but non-trivial fundamental group (hence V is not simply-connected.)
The Poincaré homology sphere therefore refutes the conjecture made at the
end of the second supplement, and it prompts the revised Poincaré conjecture
(now known to be correct): the three-sphere is the only closed three-dimensional
manifold with trivial fundamental group.
Poincaré prudently concludes the fifth supplement by remarking that inves-
tigation of the revised conjecture “would carry us too far away.”

The Poincaré conjecture

In the Analysis situs and its five supplements, Poincaré opened up a vast new
area of mathematics. It is not surprising that he left it incompletely explored.
Among the most important gaps in his coverage were:

1. The topological invariance of dimension, first proved by Brouwer (1911).

2. The topological invariance of the Betti and torsion numbers, first proved
by Alexander (1915).

3. The existence of non-homeomorphic three-dimensional manifolds with the

same fundamental group, first proved by Alexander (1919).

4. The existence of a polyhedral structure on every differentiable manifold,

first proved by Cairns (1934).

5. The existence of topological manifolds without a polyhedral structure, first

proved by Kirby and Siebenmann around 1970, and published in Kirby
and Siebenmann (1977).

But the deepest of the unsolved problems left by Poincaré was one he first
thought was trivial—the Poincaré conjecture.
In the beginning, there was no conjecture, because Poincaré thought it obvi-
ous that a simply-connected closed manifold was homeomorphic to a sphere. In
the second supplement he came up with a sharper claim that was less obvious,
hence in his view worth conjecturing: a closed manifold with trivial homol-
ogy is homeomorphic to a sphere. But on occasions thereafter he forgot that
there is a difference between trivial homology and trivial fundamental group.
Finally, the discovery of Poincaré homology sphere in Supplement 5 opened his
eyes to the real problem, and the Poincaré conjecture as we know it today was
born: a closed 3-manifold with trivial fundamental group is homeomorphic to
the 3-sphere.
10 Translator’s Introduction

The existence of homology spheres shows that three dimensions are more
complicated than two, but just how much more complicated they are was not
immediately clear. Further results on three-dimensional manifolds came with
glacial slowness, and they often revealed new complications. Dehn (1910) found
infinitely many homology spheres and Whitehead (1935) found an open three-
dimensional manifold that is simply-connected but not homeomorphic to R3 .
In the 1950s and 1960s there was at last some good news about three-
dimensional manifolds; for example, they all have a polyhedral structure (Moise
(1952)). The news did not include a proof of the Poincaré conjecture, however.
Instead, progress on the conjecture came in higher dimensions, with a proof by
Smale (1961) of the analogous conjecture for the n-dimensional sphere Sn for
n ≥ 5. Unfortunately, while three dimensions are harder than two, five are eas-
ier than three in some respects. So Smale’s proof did not throw much light on
the classical Poincaré conjecture, or on the analogous conjecture for S4 either.
The analogue of the Poincaré conjecture for four-dimensional manifolds was
finally proved by Freedman (1982). Freedman’s proof was a tour de force that si-
multaneously solved several longstanding problems about four-dimensional man-
ifolds. That his approach worked at all was a surprise to many of his colleagues,
and finding a similar approach to the classical Poincaré conjecture seemed out
of the question.
Indeed, an entirely new approach to the Poincaré conjecture had already
been taking shape in the hands of William Thurston in the late 1970s. Thurston,
like Poincaré and Dehn, was interested in geometric realizations of manifolds,
exemplified by the surfaces of constant curvature that realize all the topological
forms of closed surfaces. He conjectured that all 3-manifolds may be realized
in a similar, though more complicated, way. Instead of the three 2-dimensional
geometries of constant curvature, one has eight “homogeneous” 3-dimensional
geometries. (The eight geometries were discovered by Bianchi (1898), and re-
discovered by Thurston.) And instead of a single geometry for each 3-manifold
M one has a “decomposition” of M into finitely pieces, each carrying one of the
eight geometries.
Thurston’s geometrisation conjecture states that each closed connected 3-
manifold is homeomorphic to one with such a decomposition. The Poincaré
conjecture follows from a special case of the geometrisation conjecture for man-
ifolds of positive curvature. For more details on the evolution of the Poincaré
conjecture up to this point, see Milnor (2003).
Thurston was able to prove many cases of his geometrisation conjecture, but
geometrisation seemed to run out of steam in the early 1980s. This was not
entirely disappointing to some topologists, who still hoped for a proof of the
Poincaré conjecture by purely topological methods. However, more geometry
was to come, not less, and differential geometry at that. It was not enough to
consider manifolds with “homogeneous” geometry; one had to consider mani-
folds with arbitrary smooth geometry, and to let the geometry “flow” towards
homogeneity.
The idea of “flowing towards homogeneity” was initiated by Hamilton (1982),
using what is called the Ricci curvature flow. Hamilton was able to show that the
Comments on terminology and notation 11

Ricci curvature flow works in many cases, but he was stymied by the formation
of singularities in the general case. The difficulties were brilliantly overcome by
Grigory Perelman in 2003. Perelman published his proof only in outline, in three
papers posted on the internet in 2002 and 2003, but experts later found that
these papers contained all the ideas necessary to construct a complete proof of
the geometrisation conjecture. Perelman himself, apparently sure that he would
be vindicated, published nothing further and seems to have gone into seclusion.
For a very thorough and detailed account of Perelman’s proof of the Poincaré
conjecture, see Morgan and Tian (2007).

Comments on terminology and notation

Poincaré’s topology papers pose an unusual problem for the translator, inasmuch
as they contain numerous errors, both large and small, and misleading notation.
My policy (which is probably not entirely consistent) has been to make only
small changes where they help the modern reader—such as correcting obvious
typographical errors—but to leave serious errors untouched except for footnotes
pointing them out.
The most serious errors must be retained because they were a key stimulus to
the development of Poincaré’s thought in topology. As mentioned above, some
of the five supplements exist only because of mistakes in the Analysis situs
paper. It is more debatable whether one should retain annoying notation, such
as the + sign Poincaré uses to denote the (generally noncommutative) group
operation in the fundamental group, or the ≡ sign and the word “congruence”
he uses for the (asymmetric) boundary relation. I have opted to retain these,
partly to assist readers who wish to compare the translation with the original
papers, and also because they may be a clue to what Poincaré was thinking
when he first applied algebra to topology.
I have also retained the word “conjugate” that Poincaré uses for the paired
sides of a polyhedron, or the paired sides of a polygon, that are to be identified to
form a manifold. One can replace “conjugate” by “identified” in many cases, but
sometimes “paired” is better, so I thought it safest not to meddle. Fortunately,
Poincaré does not use the word “conjugate” in the group-theoretic sense, even
though the concept of conjugacy in group theory briefly arises.
On the other hand, I consistently use the word “manifold” where Poincaré
uses “variety,” and I call manifolds “orientable” where he calls them “two-sided”
and “non-orientable” where he calls them “one-sided.” The word “variety”
always suggests algebraic geometry today, whereas Poincaré is really thinking
about the topology of manifolds (even though many of them are in fact algebraic
varieties), so “manifold” is the right word for the modern reader. The words
“two-sided” and “one-sided” are less misleading than “variety,” but Poincaré
also uses them in a second sense, to describe separating and non-separating
curves on a surface, which have “two sides” and “one side” respectively. Calling
manifolds “orientable” and “non-orientable” therefore removes a possible source
of confusion.
12 Translator’s Introduction

Acknowledgements
I have drawn on the work of Sarkaria (1999), which gives a detailed summary,
with some comments and corrections, of Poincaré’s papers in topology. The
commentary of Dieudonné (1989) on Poincaré’s Analysis situs and its first two
supplements is also useful, as is Chapter VII of Scholz (1980).
I translated Analysis situs and the the first, second, and fifth supplements
in 1970s, when I was first learning topology. At the time, I did not think there
would be an opportunity to publish these papers, so I did not bother to translate
the remaining two supplements, which were further from my interests at the
time. Thirty years later, I was pleasantly surprised to be contacted by Andrew
Ranicki and Cameron Gordon about their classic papers project. With their
encouragement, I translated the two missing supplements, and edited the whole
sequence into the form you see today. I am delighted that Poincaré’s topological
work is finally appearing in English, and I thank Andrew and Cameron for
making this possible.
John Stillwell
Monash University and University of San Francisco
June 2009

Bibliography
Abel, N. H. (1826/1841). Mémoire sur une propriété générale d’une classe trés-étendue
de fonctions transcendantes. Mém. Acad. d. Sciences de Paris, 176–264. Received
in 1826, published in 1841.

Alexander, J. W. (1915). A proof of the invariance of certain constants of analysis

situs. Trans. Amer. Math. Soc. 16, 148–154.

Alexander, J. W. (1919). Note on two three-dimensional manifolds with the same

group. Trans. Amer. Math. Soc. 20, 339–342.

Betti, E. (1871). Sopra gli spazi di un numero qualunque di dimensioni. Annali di

Matematica pura ed applicata 4, 140–158.

Bianchi, L. (1898). Sugli spazi a tre dimensioni che ammettono un gruppo con-
tinuo di movimenti. Memorie di Matematica e di Fisica della Societa Italiana delle
Scienze 11, 267–352. English traanslation by Robert Jantzen in General Relativity
and Gravitation 33, (2001), pp. 2171–2253.

Bonnet, O. (1848). Mémoire sur la théorie générale des surfaces. J. Éc. Polytech. 19,
1–146.

Brouwer, L. E. J. (1911). Beweis der Invarianz der Dimensionzahl. Math. Ann. 70,
161–165.

Cairns, S. S. (1934). On the triangulation of regular loci. Ann. of Math. 35, 579–587.

Cayley, A. (1859). On contour and slope lines. Phil. Mag. 18, 264–268.
BIBLIOGRAPHY 13

Darboux, G. (1913/1952). Éloge Historique d’Henri Poincaré. In Œuvres de Henri

Poincaré, Tome II, pp. vii–lxxi. Paris: Gauthier-Villars.

Dehn, M. (1910). Über die Topologie des dreidimensional Raumes. Math. Ann. 69,
137–168. English translation in ?.

Dieudonné, J. (1989). A History of Algebraic and Differential Topology. 1900–1960.

Boston, MA: Birkhäuser Boston Inc.

Dyck, W. (1888). Beiträge zur Analysis situs I. Math. Ann. 32, 457–512.

Euler, L. (1752). Elementa doctrinae solidorum. Novi Comm. Acad. Sci. Petrop. 4,
109–140. In his Opera Omnia, ser. 1, 26: 71–93. English translation in ?.

Freedman, M. H. (1982). The topology of four-dimensional manifolds. J. Differential

Geom. 17, 357–453.

Gauss, C. F. (1827). Disquisitiones generales circa superficies curvas. Göttingen:

König. Ges. Wiss. Göttingen. English translation in ?.

Hamilton, R. S. (1982). Three-manifolds with positive Ricci curvature. J. Differential

Geom. 17, 255–306.

Heegaard, P. (1898). Forstudier til en topologisk Teorie for de algebraiske Fladers Sam-
menhæng. Copenhagen: University of Copenhagen Dissertation. French translation:
Sur l’Analysis situs in Bull. Math. Soc. France 44 (1916), pp. 161–242.

Kirby, R. C. and L. C. Siebenmann (1977). Foundational Essays on Topological Mani-

folds, Smoothings, and Triangulations. Princeton, N.J.: Princeton University Press.

Klein, F. (1882). Neue Beiträge zur Riemanschen Funktionentheorie. Math. Ann. 21,
141–218. In his Gesammellte Mathematische Abhandlungen 3: 630–710.

Maxwell, J. C. (1870). On hills and dales. Phil. Mag. 40, 421–427.

Milnor, J. (2003). Towards the Poincaré conjecture and the classification of 3-

manifolds. Notices Amer. Math. Soc. 50, 1226–1233.

Möbius, A. F. (1863). Theorie der Elementaren Verwandtschaft. Werke 2: 433-471.

Moise, E. E. (1952). Affine structures in 3-manifolds. V. The triangulation theorem

and Hauptvermutung. Ann. of Math. (2) 56, 96–114.

Morgan, J. and G. Tian (2007). Ricci Flow and the Poincaré Conjecture, Volume 3 of
Clay Mathematics Monographs. Providence, RI: American Mathematical Society.

Noether, E. (1926). Ableitung der Elementarteilertheorie aus der Gruppentheorie.

Jber. Deutsch. Math. Verein. 36, 104.

Picard, E. (1889). Mémoire sur la théorie des fonctions algébriques de deux variables.
J. de Math. 5, 135–319.

Poincaré, H. (1892). Sur l’analysis situs. Comptes rendus de l’Academie des Sci-
ences 115, 633–636.
14 Translator’s Introduction

Poincaré, H. (1895). Analysis situs. J. Éc. Polytech., ser. 2 1, 1–123.

Poincaré, H. (1985). Papers on Fuchsian Functions. New York: Springer-Verlag.

Translated from the French and with an introduction by John Stillwell.

Riemann, G. F. B. (1851). Grundlagen für eine allgemeine Theorie der Functionen

einer veränderlichen complexen Grösse. Werke, 2nd ed., 3–48.

Sarkaria, K. S. (1999). The topological work of Henri Poincaré. In History of Topology,

pp. 123–167. Amsterdam: North-Holland.

Scholz, E. (1980). Geschichte des Mannigfaltigkeitsbegriffs von Riemann bis Poincaré.

Mass.: Birkhäuser Boston.

Smale, S. (1961). Generalized Poincaré conjecture in dimensions greater than four.

Ann. of Math. 74, 391–406.

Whitehead, J. H. C. (1935). A certain open manifold whose group is unity. Quart. J.

Math. 6, 268–279.
On Analysis situs 15

ON ANALYSIS SITUS
Comptes rendus de l’Académie des Sciences 115 (1892), pp. 633-636.

One knows what is meant by the connectivity of a surface, and the important
role this notion plays in the general theory of functions, despite being borrowed
from an entirely different branch of mathematics, namely the geometry of situ-
ation or Analysis situs.
It is because researches of this kind can have applications outside geometry
that it is of interest to pursue them to spaces of more than three dimensions.
Riemann understood this well; and he wished to extend his beautiful discovery
and apply it to the Analysis situs of general spaces, but unfortunately he left
the subject in a very incomplete state. Betti, in volume IV, series 2 of Annali
di Matematica, recovered and completed Riemann’s results. He considered a
surface (manifold of dimension n) in the space of n + 1 dimensions, and defined
n − 1 numbers
p1 , p2 , . . . , pn−1
that he called he called the orders of connection of the surface.
Persons who recoil from geometry of more than three dimensions may believe
this result to be useless and view it as a futile game, if they have not been
informed of their error by the use made of Betti numbers by our colleague M.
Picard in pure analysis and ordinary geometry.
Meanwhile, the field is by no means exhausted. One may ask whether the
Betti numbers suffice to determine a closed surface from the viewpoint of Anal-
ysis situs. That is, given two surfaces with the same Betti numbers, we ask
whether it is possible to pass from one to the other by a continuous deforma-
tion. This is true in the space of three dimensions, and we may be inclined to
believe that it is again true in any space. The contrary is true.
In order to explain, I want to approach the question from a new viewpoint.
Let x1 , x2 , . . . , xn+1 be the coordinates of a point on the surface. These n + 1
quantities are connected by the equation of the surface. Now let

F1 , F2 , ..., Fp

be any p functions of the n + 1 coordinates x (which I always suppose to be

connected by the equation of the surface, and which I suppose to take only real
values).
I do not assume that the functions F are uniform, but I suppose that if the
point (x1 , x2 , . . . , xn+1 ) describes an infinitely small contour on the surface then
each of the functions F returns to its initial value. This being so, we suppose
that our point now describes a finite closed contour on the surface. It may then
happen that the p functions do not return to their initial values, but instead
become
F10 , F20 , . . . , Fp0 .
16 On Analysis Situs

In other words, they undergo the substitution

(F1 , F2 , . . . , Fp ; F10 , F20 , . . . , FP0 ).

All the substitutions corresponding to the different closed contours that we

can trace on the surface form a group which is discontinuous (at least as far as
its form is concerned).
This group evidently depends on the choice of functions F . We suppose first
that these functions are the most one can imagine, other than being subject to
the condition imposed above, and let G be the corresponding group. If G0 is the
group corresponding to another choice of functions, then G0 will be isomorphic
to G—holoedrically in general but meriedrically in special cases.1
The group G can then serve to define the form of the surface and it is called
the group of the surface.2 It is clear that if two surfaces can each be transformed
to the other by a continuous transformation, then their groups are isomorphic.
The converse, though less evident, is again true for closed surfaces, so that what
defines a closed surface, from the viewpoint of Analysis situs, is its group.3
This leads us to pose the following question: do two surfaces with the same
Betti numbers always have the same group?
To resolve this question we make use of a simple mode of representation
in ordinary space when we want to define a surface in four-dimensional space.
We consider a properly discontinuous group G in ordinary space. The space is
thereby decomposed into infinitely many fundamental domains, each the trans-
form of some other by one of the transformations in the group. I suppose that
the fundamental domain does not extend to infinity and that each [non-identity]
substitution in the group has no fixed point.
Let
X1 , X2 , X3 , X4
be four functions of the coordinates x, y, z of ordinary space that are invariant
under substitutions in the group G. If we consider X1 , X2 , X3 , X4 as the coordi-
nates of a point in four-dimensional space, this point describes a closed surface
whose group is isomorphic to G, and holoedrically so if the functions X are the
most general possible among those that are invariant under G.
We consider, in particular, the group generated by the three substitutions

(x, y, z; x + 1, y, z),
(x, y, z; x, y + 1, z),
(x, y, z; αx + βy, γx + δy, z + 1),
1 Here Poincaré is using the 19th-century terminology, where a “holoedric isomorphism” is

what we call an isomorphism, and a “meriedric isomorphism” is a homomorphism. (Transla-

tor’s note.)
2 Later called the fundamental group by Poincaré in §12 of his Analysis situs paper. (Trans-

lator’s note.)
3 This is true for surfaces in the traditional, two-dimensional, sense, but not for manifolds of

three dimensions. Near the end of §14 of Analysis situs, Poincaré raised the question whether
two manifolds with the same group are necessarily homeomorphic. (Translator’s note.)
On Analysis situs 17

where α, β, γ, δ are integers such that αδ − βγ = 1. I call this the group

(α, β, γ, δ) for short.
It has a cube as fundamental domain.4
We first observe that the two groups (α, β, γ, δ) and (α0 , β 0 , γ 0 , δ 0 ) cannot be
isomorphic unless the two transformations

(x, y; αx + βy, γx + δy), (x, y; α0 x + β 0 y, γ 0 x + δ 0 y)

are transforms5 of each other by a linear transformation with integer coefficients.

This does not happen in general.
We now seek to determine the Betti numbers for the surface with group
(α, β, γ, δ). We see that one of the orders of connectivity is always quadruple
and the other is

double in the general case;

triple if α + δ = 2 :
quadruple if α = δ = 1, β = γ = 0.

It follows that the Betti numbers can be the same for two surfaces without
their groups being isomorphic and, consequently, without it being possible to
pass from one surface to the other by a continuous deformation.
This remark throws some light on the theory of ordinary algebraic surfaces
and makes less strange the discovery of M. Picard, according to which the
surfaces have no one-dimensional cycle if they are the most general of their
degree.

4 Moreover, we see that the vertical sides of the cube are mapped onto each other by

unit translations in the x- and y-directions, so that each horizontal cross-section of the cube
becomes a torus. The top and bottom faces of the cube are also mapped onto each other (by
the more complicated third generator of G), so that the manifold obtained is what we would
now call a torus bundle over the circle. (Translator’s note.)
5 What we would now call conjugates. (Translator’s note.)
18 Analysis Situs

ANALYSIS SITUS
Journal de l’École Polytechnique 1 (1895), pp. 1-121.

Introduction
Nobody doubts nowadays that the geometry of n dimensions is a real object.
Figures in hyperspace are as susceptible to precise definition as those in ordinary
space, and even if we cannot represent them, we can still conceive of them
and study them. So if the mechanics of more than three dimensions is to be
condemned as lacking in object, the same cannot be said of hypergeometry.
Geometry, in fact, has a unique raison d’être as the immediate description
of the structures which underlie our senses; it is above all the analytic study of
a group; consequently there is nothing to prevent us proceeding to study other
groups which are analogous but more general.
But why, it may be said, not preserve the analytic language and replace
the language of geometry, as this will have the advantage that the senses can
no longer intervene. It is that the new language is more concise; it is the
analogy with ordinary geometry which can create fruitful associations of ideas
and suggest useful generalizations.
Perhaps these reasons are not sufficient in themselves? It is not enough, in
fact, for a science to be legitimate; its utility must be incontestable. So many
objects demand our attention that only the most important have the right to
be considered.
Also, there are parts of hypergeometry which do not have a place of great
interest: for example, researches on the curvature of hypersurfaces in the space
of n dimensions. We are certain in advance of obtaining the same results as in
ordinary geometry, and we need not undertake a long voyage to view a spectacle
like the one we encounter at home.
But there are problems where the analytic language is entirely unsuitable.
We know how useful geometric figures are in the theory of imaginary func-
tions and integrals evaluated between imaginary limits, and how much we desire
their assistance when we want to study, for example, functions of two complex
variables.
If we try to account for the nature of this assistance, figures first of all make
up for the infirmity of our intellect by calling on the aid of our senses; but
not only this. It is worthy repeating that geometry is the art of reasoning well
from badly drawn figures; however, these figures, if they are not to deceive us,
must satisfy certain conditions; the proportions may be grossly altered, but the
relative positions of the different parts must not be upset.
The use of figures is, above all, then, for the purpose of making known
certain relations between the objects that we study, and these relations are
those which occupy the branch of geometry that we have called Analysis situs,
and which describes the relative situation of points and lines on surfaces, without
consideration of their magnitude.
Introduction 19

The fact that relations of the same nature hold between the objects of hyper-
surface, so that there is then an Analysis situs of more than three dimensions,
is due, as we have shown, to Riemann and Betti.
This science enables us to know the nature of these relations, although this
knowledge is less intuitive, since it lacks a counterpart in our senses. Indeed, in
certain cases it renders us the service that we ordinarily demand of geometrical
figures.
I shall confine myself to three examples.
The classification of algebraic curves into types rests, after Riemann, on the
classification of real closed surfaces from the point of view of Analysis situs. An
immediate induction shows us that the classification of algebraic surfaces and
the theory of their birational transformations are intimately connected with the
classification of real closed hypersurfaces in the space of five dimensions from
the point of view of Analysis situs. M. Picard, in a memoir honoured by the
Académie des Sciences, has already insisted on this point.
Then again, in a series of memoirs in Liouville’s journal, entitled: Sur les
courbes définies par les équations différentielles I have employed the ordinary
analysis situs of three dimensions in the study of differential equations. The
same researches have been pursued by M. Walther Dyck. We can easily see that
generalized Analysis situs will permit us to treat higher order equations in the
same way, in particular, the equations of celestial mechanics.
M. Jordan?? has determined analytically the groups of finite order contained
in the linear group of n variables. Before that, M. Klein had resolved the same
problem for the linear group of two variables, by a geometric method of rare
elegance. Could not the method of M. Klein be extended to the group of n
variables, or any continuous group? I have not been able to succeed so far,
but I have thought a great deal about the question and it seems to me that the
solution must depend on a problem of Analysis situs and that the generalization
of celebrated Euler polyhedron theorem must play a rôle.
I do not think then that I have engaged in useless work in writing the present
memoir; I regret only that it is too long, but when I try to restrict myself I fall
into obscurity; I prefer to be considered a little loquacious.
20 Analysis Situs

Table of Contents

§1. First definition of manifold 20

§2. Homeomorphism 22
§3. Second definition of manifold 24
§4. Oppositely oriented manifolds 28
§5. Homologies 30
§6. Betti numbers 30
§7. Use of integrals 32
§8. Orientable and non-orientable manifolds 34
§9. Intersection of two manifolds 40
§10. Geometric representation 49
§11. Representation by a discontinuous group 54
§12. The fundamental group 58
§13. Fundamental equivalences 60
§14. Conditions for homeomorphism 65
§15. Other modes of generation 75
§16. The theorem of Euler 85
§17. The case where p is odd 93
§18. Second proof 95

§1. First definition of manifold

Let x1 , x2 , . . . , xn be n variables, which can be regarded as the coordinates of a

point in n-dimensional space.
For the time being I assume that these n variables are always real.
Any sequence of n variables will be called a point. We consider the following
system consisting of p equations and q inequalities.



 F1 (x1 , x2 , . . . , xn ) = 0



 F2 (x1 , x2 , . . . , xn ) = 0



 ...

Fp (x1 , x2 , . . . , xn ) = 0
(1)
 ϕ1 (x1 , x2 , . . . , xn )
 > 0



 ϕ2 (x1 , x2 , . . . , xn ) > 0



 ...

ϕq (x1 , x2 , . . . , xn ) > 0

I assume that the functions F and ϕ are uniform and continuous and that
they have continuous derivatives; in addition I assume that if we form the matrix
§1. First definition of manifold 21

∂F1 ∂F1 ∂F1

∂x1 ∂x2 ... ∂xn

∂F2 ∂F2 ∂Fn

∂x1 ∂x2 ... ∂xn

... ... ... ...

∂Fp ∂Fp ∂Fp

∂x1 ∂x2 ... ∂xn

and form the determinants obtained by taking any p columns, then these deter-
minants are never all zero simultaneously.
I shall say that the set of points which satisfies the conditions (1) constitutes
a manifold of n − p dimensions. If in particular p = 0, so that there are no
equations, I have an n-dimensional manifold which is nothing but a portion of
the n-dimensional space; ordinarily I shall describe this manifold as a domain.
Two cases can occur. Let α1 , α2 , . . . , αn ; β1 , β2 , . . . , βn be two points satis-
fying the conditions (1). Then either it is possible to vary x1 , x2 , . . . , xn in a
continuous manner from α1 , α2 , . . . , αn to β1 , β2 , . . . , βn without violating the
conditions (1), for any values α1 , α2 , . . . , αn and β1 , β2 , . . . , βn which satisfy
these conditions, or else this is not always possible.
In the first case we say that the manifold defined by the conditions (1) is
connected.
In what follows I shall generally consider connected manifolds, and, to the
extent that non-connected manifolds concern us, I confine myself to observing
that they can always be decomposed into a finite or infinite number of connected
manifolds.
Consider for example the manifold

x22 + x41 − 4x21 + 1 = 0.

Here n = 2 and the point x1 , x2 is a point of the plane; our manifold is then
none other than a 4th degree curve; however, since that curve is composed of
two closed branches, our manifold is not connected.
But we can decompose it into two others, namely

x22 + x41 − 4x21 + 1 = 0 x1 > 0

and
x22 + x41 − 4x21 + 1 = 0 x1 < 0
and each of these, being a single closed branch of the curve, is connected.
I shall say that a manifold is finite if all its points satisfy the condition

x21 + x22 + · · · + x2n < K 2

where K is a given constant.

22 Analysis Situs

We now consider the system of relations



 Fα = 0 (α = 1, 2, . . . , p)
(2) ϕβ = 0

 ϕ >0 >
γ (γ < β)

consisting of p + 1 equations and q − 1 inequalities.

It can happen that there is no point satisfying the conditions (2), or there
may be, in which case such points constitute a manifold of less than n − p
dimensions.
The set of points which satisfy one of the q systems of relations

>

 Fα = 0, ϕ1 = 0, ϕγ > 0 (γ < 1)


 >
(3) Fα = 0, ϕ2 = 0, ϕγ > 0 (γ < 2)

 ... ... ... ...


 >
Fα = 0, ϕq = 0, ϕγ > 0 (γ < q)

is called the boundary of the manifold defined by the conditions (1). However,
we shall sometimes take another point of view and only consider those which
have n − p − 1 dimensions as true boundaries.
It may happen that there is no manifold of n − p − 1 dimensions satisfying
any of the q systems (3). In that case, the manifold defined by the conditions
(1) will be called unbounded. In the contrary case it will be called bounded.
If a manifold is simultaneously finite, connected and bounded, it will be
called closed.
To abbreviate our language a little we shall give the name (hyper)surfaces
to manifolds of n − 1 dimensions, except for the case n = 2, in which case we
give them the name curves.

§2. Homeomorphism
Consider a substitution which changes x1 , x2 , . . . , xn into x01 , x02 , . . . , x0n , subject
only to the following conditions.
We have

(4) x0i = ϕi (x1 , x2 , . . . , xn ) (i = 1, 2, . . . , n).

In a certain domain the functions ϕi are uniform, finite and continuous; they
have continuous derivatives and their Jacobian is non-zero.
If we solve the equations (4) for x1 , x2 , . . . , xn we get

xk = ϕ0k (x01 , x02 , . . . , x0n ) (k = 1, 2, . . . , n)

§2. Homeomorphism 23

and the functions ϕ0k satisfy the same conditions as the functions ϕi .
It is clear that the set of substitutions which satisfy these conditions consti-
tutes a group, and this group is one of the most general that we can imagine.
The science whose object is the study of this group and its analogues receives
the name Analysis situs.
It is clear that a substitution of the group transforms a manifold of m di-
mensions into a manifold of m dimensions, and that the new manifold will be
connected, or finite, or unbounded according as this is the case for the original
manifold (and conversely).
Consider two manifolds V and V 0 of the same number of dimensions defined
by the respective conditions
½
Fα = 0 (α = 1, 2, . . . , p)
(1)
ϕβ > 0 (β = 1, 2, . . . , q)

and
½
Fα0 = 0 (α = 1, 2, . . . , p)
(10 )
ϕ0β > 0 (β = 1, 2, . . . , q)

Suppose that we can make a point x1 , x2 , . . . , xn of the manifold V correspond

to a point x01 , x02 , . . . , x0n of the manifold V 0 , in such a way that we have

(5) x0k = ψk (x1 , x2 , . . . , xn ) (k = 1, 2, . . . , n).

I consider the domain D defined by the inequalities

Fα > −ε, Fα < ε, ϕβ > 0.

The manifold V is evidently contained entirely in the domain D.

I assume that in the domain D the functions ψk are finite, continuous and
uniform, that they have continuous derivatives and that their Jacobian is non-
zero.
Solving the equations (5) we find

(6) xk = ψk0 (x01 , x02 , . . . , x0n ) (k = 1, 2, . . . , n).

I consider the domain D0 defined by the inequalities

Fα0 > −ε, Fα0 < ε, ϕ0β > 0,

and I assume that in the domain D0 the functions ψk0 are finite, continuous
and uniform, that they have continuous derivatives and that their Jacobian is
non-zero.
It follows from these hypotheses that each point of V corresponds to exactly
one point of V 0 , and conversely; to every manifold W contained in V there
corresponds a manifold W 0 , of the same number of dimensions, contained in V 0 ;
if W is connected, finite or unbounded the same is true of W 0 and conversely.
24 Analysis Situs

If all these conditions are satisfied we say that the two manifolds V and V 0
are equivalent from the point of view of Analysis situs, or simply homeomorphic,
that is to say, of the same form.
I may also say that two more complicated figures, composed of any number
of manifolds, are homeomorphic when one passes into the other by a transfor-
mation of the form (5).
Thus two polygons with the same number of sides are homeomorphic, two
polyhedra with the same arrangement of faces and the same number of sides on
each face are homeomorphic, etc..

§3. Second definition of manifold

One can define manifolds in an entirely different manner. Consider n equations



 x1 = θ1 (y1 , y2 , . . . , ym )

x2 = θ2 (y1 , y2 , . . . , ym )
(8)

 .....................

xn = θn (y1 , y2 , . . . , ym )

It is clear that these equations (if the y are regarded as independent variables)
represent a manifold of m dimensions.
We again have a manifold of m dimensions if we adjoin to the equations (8)
a certain number of inequalities of the form

ψ(y1 , y2 , . . . , ym ) > 0

which limit the field of variation of the variables y.

I assume that the functions θ are finite and continuous; but I shall make a
further hypothesis which does lessen generality in an essential fashion (moreover,
this could have been done likewise with my first definition of manifold).
I shall suppose, namely, that the functions θ are analytic. If, in fact, the
functions θ are finite and continuous, we can find functions θ0 which are analytic
and differ from the θ by as little as we wish.
I assume, in addition, that the Jacobians of the m functions θ with respect
to the y are never simultaneously zero.
We obtain a manifold no different from the first by replacing the y in the
equations (8) by any m analytic functions of the m variables z1 , z2 , . . . , zm .
The power of this new definition would be limited if it were not for the fact
that we could augment by the procedure of analytic continuation.
We consider two manifolds V and V 0 defined by equations analogous to (8).
For example, let the equations

(8) xi = θi (y1 , y2 , . . . , ym )
§3. Second definition of manifold 25

define V and let the equations

(80 ) xi = θi0 (y1 , y2 , . . . , ym )

define V 0 .
It can happen that the two manifolds have a common part V 00 also of m
dimensions.
In that case, in the interior of V 00 , the y will be analytic functions of the y 0
and conversely.
We then say that the two manifolds V and V 0 are analytic continuations of
each other.
In this way we can form a chain of manifolds

V1 , V2 , ..., Vn

such that each is an analytic continuation of its predecessor, and there is a

common part between any two consecutive manifolds of the chain. I shall call
this a connected chain.
It can also happen that the chain is closed, i.e. that Vn is the same as V1 .
We can then have a network of manifolds, that is to say, a set of manifolds
in which each is the continuation of certain others and we can pass from any
one to any one by analytic continuation; I shall call this a connected network.
We could then consider the set of all manifolds of the same chain or the same
network as forming a unique manifold.
This is a broader definition than the first one.
There are, in fact, manifolds (and we shall see examples later) which can
be decomposed into a certain number of partial manifolds forming a connected
chain or network and such that each of them can be defined by equations of the
form (8) (manifolds which, consequently, are covered by our second definition),
which nevertheless cannot be defined by relations of the form (1) and hence are
not covered by our first definition.
On the other hand, each manifold which satisfies the first definition also
satisfies the second.
In fact, by a well-known theorem, if y1 , y2 , . . . , yn are defined by n relations
of the form

 y1 = F1 (x1 , x2 , . . . , xn )


y2 = F2 (x1 , x2 , . . . , xn )
(α)

 .....................

yn = Fn (x1 , x2 , . . . , xn )

and if, in a certain domain, the F are holomorphic functions of the x, with
non-zero Jacobian, then we can solve the equations for the

(β) xi = θi (y1 , y2 , . . . , yn )

where the θi are holomorphic functions of the y.

26 Analysis Situs

Now consider a manifold V satisfying our first definition, i.e. defined by

relations

(1) Fα = 0, ϕβ > 0.

Referring to the equations (α), we take for F1 , F2 , . . . , Fp the first members of

the p equations (1). As regards the other functions

Fp+1 , Fp+2 , ..., Fn

– we take any n − p holomorphic functions of the x. I subject them to only one

condition.
Let x01 , x02 , . . . , x0n be any point M0 of the manifold V . I arrange the Jacobian
of the n functions F in such a way that it does not vanish for

xi = x0i .

This is evidently possible, since I assumed that the Jacobians of the p func-
tions
F 1 , F2 , . . . , Fp
with respect to any p of the variables x do not simultaneously vanish.
I can also assume that Fp+1 , Fp+2 , . . . , Fn vanish at the point M0 , i.e. for

xi = x0i .

Then, by the theorem cited above, we can solve the equations (α) and we
find that the xi ’s are expressible in series of powers of

y1 , y2 , ..., yn ,

convergent when these quantities satisfy certain inequalities.

Then let

(β) xi = θi

be these equations and let

λk (y1 , y2 , . . . , yn ) > 0

be the conditions the y must satisfy to guarantee convergence of the series.

If we now make
y1 = y2 = · · · = yp = 0,
the initial p equations (α) are the same as the p equations (1); and the functions
θi , not dependent on more than n − p = m variables, are of the form (8).
Then the set of relations

xi = θi , ϕβ > 0, λk > 0
§3. Second definition of manifold 27

represents a manifold v defined in the second manner which is the same as the
part of V defined by
Fα = 0, ϕβ > 0, λk > 0.
The point M0 , which is an arbitrary point of V , is part of v. We can then
construct, around any point of V , a manifold analogous to v.
The simplest case is where the convergence conditions λk > 0 are conse-
quences of the inequalities ϕβ > 0. Then v is the same as V , and to define it
we can be content with the equations (8) and the inequalities
(9) ψβ > 0.
We remark in passing that the Jacobians of m of the functions θ with respect
to the y are not simultaneously zero.
If the conditions λβ > 0 are not consequences of the inequalities ϕβ > 0, we
decompose the manifold V into partial manifolds; thus for example when ψ is
any function of x1 , x2 , . . . , xn the manifold V can evidently be decomposed into
two partial manifolds
Fα = 0, ϕβ > 0, ψ > 0
and
Fα = 0, ϕβ > 0, ψ < 0.
Given that we can always decompose V into partial manifolds, or better,
construct a number of partial manifolds on V which overlap each other to an
arbitrarily small extent, we find for each of them a system of auxiliary variables
which permit that partial manifold to be represented by equations and inequal-
ities of the form (8) and(9) satisfying all the conditions enunciated above. Each
point M0 of V belongs to one of these partial manifolds and the set of these
manifolds forms a connected net. Thus the first definition is restored to the
second.
Nevertheless, it remains to remark that it can happen that two different
systems of values of y1 , y2 , . . . , yn correspond to the same system of values of
the functions θ1 , θ2 , . . . , θn and hence to the same point of the manifold V .
In that case it is convenient to adjoin to the inequalities (9) the further
inequalities
(10) ψγ > 0
chosen so that different systems of values of the variables y which correspond
to the same point of V always satisfy one and only one of the inequalities (9)
and (10).
Take for example a torus, with equation
(x21 + x22 + x23 + R2 − r2 )2 − 4R2 (x21 + x22 ) = 0
We set
x1 = (R + r cos y1 ) cos y2
x2 = (R + r cos y1 ) sin y2
x3 = r sin y1
28 Analysis Situs

and we see that the same point of the torus corresponds to an infinity of systems
of values of the y, comprised by the formulae

y1 + 2K1 π, y2 + 2K2 π

where the K are integers.

But if we constrain the y by the conditions

0 ≤ y1 < 2π, 0 ≤ y2 < 2π

we shall have only one system of values y corresponding to each point of the
torus.

§4. Oppositely oriented manifolds

With each definition there is reason to make a distinction, the importance
of which will be realized later.
Suppose firstly that we have a manifold V defined in the first manner, i.e.
by relations
Fα = 0, ϕβ > 0.
We take account of the order in which the equations Fα = 0 are arranged; if
two of the equations are permuted it will be convenient to say that the system of
relations represents, not the manifold V , but the oppositely oriented manifold.
We can now replace the equations

Fα = 0 (α = 1, 2, . . . , p)

by the following:

Φ1 = A11 F1 + A12 F2 + · · · + A1p Fp = 0

Φ2 = A21 F1 + A22 F2 + · · · + A2p Fp = 0

.....................
Φp = Ap1 F1 + Ap2 F2 + · · · + App Fp = 0
where the Aik are any functions of the x.
If the determinant ∆ of the coefficients Aik does not vanish in the domain
considered, the equations Φα = 0 will be equivalent to the equations Fα = 0
and consequently, if we adjoin to them a certain number of inequalities they will
again represent the manifold V or the oppositely oriented manifold.
We shall agree to say that if ∆ (which always has the same sign, since I
have assumed it does not vanish) is positive then these equations represent
the manifold V , and if ∆ is negative, they represent the oppositely oriented
manifold.
§4. Oppositely oriented manifolds 29

If we imagine replacing the y by m analytic functions of m new variables

z1 , z2 , . . . , zm , so that we have

xi = θi0 (z1 , z2 , . . . , zm )

then the new equations again represent V or the oppositely oriented manifold.
We have to assume that the Jacobian ∆ of the y with respect to the z never
vanishes, so that a system of values of y corresponds to only a single system of
values of z. It then will always have the same sign.
We shall agree to say that if ∆ is positive the new equations again represent
V , and if it is negative they represent the oppositely oriented manifold.
We now see what happens when we pass from one definition to the other.
Let a manifold V be defined by

F1 = F2 = · · · = Fp = 0

and certain inequalities.

We adjoin to these p equations the following:

y1 = Fp+1 , y2 = Fp+2 , . . . , yn−p = Fn

where Fp+1 , y2 = Fp+2 , . . . , yn−p = Fn are any n − p functions of the x.

We have seen that if the Jacobian ∆ of the n functions F1 , F2 , . . . , Fn does
not vanish we can solve the n equations for the x and thus find n equations

xi = θi (y1 , y2 , . . . , yn−p )

which represent a manifold of n − p dimensions. But we can ask whether this

represents V or the oppositely oriented manifold.
We agree to say that it represents V if ∆ is positive, and the oppositely
oriented manifold if ∆ is negative.
Consider the manifold of n − p dimensions

Fα = 0, . . . , ϕβ > 0

which I call V . We have seen that the manifold

Fα = 0, ϕγ = 0, ϕβ > 0 (β 6= γ)

of n − p − 1 dimensions, which I call v, forms part of the boundary of V .

However, it is important to arrange the equations which define v in the
following order

F1 = 0, F2 = 0, . . . , Fp = 0, ϕγ = 0

since if two of them are permuted, we have agreed to say that these equations
no longer represent the boundary of V , or a part of it, but a manifold oppositely
oriented to that boundary.
30 Analysis Situs

§5. Homologies
Consider a manifold V of p dimensions; now let W be a manifold of q dimensions
(q ≤ p) which is a part of V . We suppose that the boundary of W is composed
of λ manifolds of q − 1 dimensions

v1 , v2 , ..., vλ .

We express this fact by the notation

v1 + v2 + · · · + vλ ∼ 0

More generally, the notation

k1 v1 + k2 v2 ∼ k3 v3 + k4 v4

where the k are integers and the v are manifolds of q − 1 dimensions will denote
that there exists a manifold W of q dimensions forming part of V , the boundary
of which is composed of k1 manifolds similar to v1 , k2 manifolds similar to v2 ,
k3 manifolds similar to v3 but oppositely oriented, and k4 manifolds similar to
v4 but oppositely oriented.
Relations of this form will be called homologies.
Homologies can be combined like ordinary equations. Thus we employ the
following notation; assuming we have

k1 v1 + k2 v2 + · · · + kp vp ∼ w1 + w2 + · · · + wp

and that the manifolds w1 , w2 , . . . , wp form part of the boundary of V ; we shall

occasionally write
k1 v1 + k2 v2 + kp vp ∼ ε.

§6. Betti numbers

We say that the manifolds

v1 , v2 , ..., vλ

which have the same number of dimensions and form part of V are linearly
independent if they are not connected by any homology with integral coefficients.
If there exist Pm − 1 closed manifolds of m dimensions which are linearly
independent and form part of V , but not more than Pm − 1, then we shall say
that the connectivity of V with respect to manifolds of m dimensions is equal
to Pm .
§6. Betti numbers 31

Thus for a manifold V of m dimensions there are m − 1 numbers which I

call
P1 , P2 , ..., Pm−1
and which are the connectivities of V with respect to manifolds of
1, 2, ..., m−1
dimensions.
I shall call this the sequence of Betti numbers in what follows.
The definitions may be clarified by an example:
Let D be a domain forming part of ordinary space and bounded by n closed
surfaces
S1 , S2 , . . . , S n
which do not intersect.
This domain is a manifold of three dimensions. It then admits two Betti
numbers, P1 and P2 .
This manifold is defined by the inequalities
ϕ1 > 0, ϕ2 > 0, . . . , ϕn > 0
if the equations to the n surfaces S are
ϕ1 = 0, ϕ2 = 0, . . . , ϕn = 0.
Since the surfaces do not intersect, there are no values x1 , x2 , x3 which si-
multaneously satisfy two of these equations
ϕi = 0, ϕk = 0.
Since the surfaces S1 , S2 , . . . , Sn are two-dimensional, they each have only a
single Betti number, which will be the Riemann connectivity; let
2Q1 + 1, 2Q2 + 1, ..., 2Qn + 1
be the connectivities (which are odd, because the surfaces are closed) of the n
surfaces
S 1 , S 2 , . . . , Sn .
Then we have
P2 = n, P1 = Q1 + Q2 + · · · + Qn + 1.
Thus, for the region inside a sphere
P2 = 1, P1 = 1
for the region inside two spheres
P2 = 2, P1 = 1
for the region inside a torus
P2 = 1, P1 = 2
for the region inside two tori
P2 = 2, P1 = 2.
32 Analysis Situs

§7. The use of integrals

Consider a manifold V which we may represent by the equations and in-
equalities (8), (9) and (10) in such a way that all the conditions enunciated
above are satisfied.
We know then what we mean by an m-tuple multiple integral
Z
F dy1 dy2 . . . dym

over the manifold V ; F of course denotes a given function of the y. Integration

can be effected with respect to the m variables successively, and the limits of
integration are defined by the inequalities (9) and (10).
That being given, I am going to define the following integral
Z X
(11) Xα1 α2 ...αm dxα1 dxα2 . . . dxαm .

The differentials dxα1 , dxα2 , . . . , dxαm are any m of the n differentials dx1 , dx2 ,
. . . , dxn . The functions Xα1 α2 ...αm are given functions of x1 , x2 , . . . , xn having
all possible combinations of the indices

α1 , α2 , . . . , αm ,

that is to say, all combinations of n letters m at a time. We make the convention

that the function X is zero if two of the indices are equal, and changes sign when
two of the indices are permuted.
That being given, the integral (11) by definition will equal the integral of
order m Z X
∂(xα1 , xα2 , . . . , xαm )
Xα1 α2 ...αm dy1 dy2 . . . dym .
∂(y1 , y2 , . . . , ym )
Now if the manifold V is not susceptible to representation by relations of
the form (8), (9) and (10) satisfying all the conditions previously enunciated, we
decompose the manifold V into partial manifolds small enough to be susceptible
to that mode of representation, and the integral (11), understood to be over the
total manifold V will, by definition, be the sum of the integrals (11), understood
to be over the various partial manifolds.
This definition may nevertheless still be ambiguous.
In fact, if we permute two of the letters y1 and y2 the integral changes sign;
it is important then to order these letters and make a permutation of two of
these letters equivalent to a change of the sense of integration in the study of
simple integrals. I shall speak then of the sense of integration in connection with
the order in which we find it convenient to arrange the letters y1 , y2 , . . . , ym .
I had occasion to deal with an analogous question in a memoir Sur les résidus
des intégrales doubles in vol. IX of Acta Mathematica, in particular in the
paragraph 3 of that memoir entitled: Conditions d’ intégrabilité.
§7. The use of integrals 33

I investigated under what circumstances these conditions of integrability are

complete, i.e. under what circumstances the integral (11) is always zero when
taken over a closed manifold.
Here is what I found; writing

(α1 , α2 , . . . , αm )

in place of Xα1 α2 ...αm and [αp ] in place of xαp , the conditions of integrability
may be written

 ∂(α1 ,α2 ,...,αm ) ∂(α2 ,α3 ,...,αm−1 )


 ∂[αm+1 ] ± ∂[α1 ]
(12)

 ± ∂(α3 ,α4 ,...,αm ,αm+1 ,α1 ) ± · · · ± ∂(αm+1 ,α2 ,...,αm−1 )
∂[α2 ] ∂[αm ] = 0

Here the following law governs the choice of the signs ±. We always take the
sign + if m is even, and alternate + and − if m is odd.
We have as many equations (12) as there are systems of indices

α1 , α2 , ..., αm , αm+1

i.e., since the indices may be chosen from the letters

1, 2, ..., n

as many as the number of combinations of n letters m + 1 at a time.

We suppose now that the conditions (12), instead of being satisfied for all
possible values of the n variables

x1 , x2 , ..., xn

are satisfied only for certain values of these variables. For example, consider a
manifold V defined by the conditions

Fα = 0, ϕβ > 0.

Next, let a domain D be defined containing all points near to V , for example,
by the conditions
−ε < Fα < ε, ϕβ > −ε
where ε is a small positive number.
We assume that the conditions (12) are satisfied for all points of the domain
D.
By repeating the reasoning of the memoir cited, with suitable modifications,
we can show the following: let a manifold V 0 of m+1 dimensions form part of V
(the number m + 1 must then be less than or equal to the number of dimensions
of V ). We assume that the boundary of V 0 is composed of k manifolds of m
dimensions
W 1 , W 2 , . . . , Wk
34 Analysis Situs

so that W1 + W2 + · · · + Wk ∼ 0.
Then if the integral (11) satisfies the conditions (12) in the domain D, the
algebraic sum of the integrals (11) taken over the manifolds W1 , W2 , . . . , Wk is
zero. It is necessary, of course, to pay attention to the sense of integration for
each of them.
The conditions (12) are sufficient for this to happen, but they are not neces-
sary; these conditions, as we have seen, are equal in number to the combinations
of n letters m + 1 at a time; it would suffice for the integral (11) to satisfy, at
all points of V , certain conditions equal in number to the combinations of n − p
letters m + 1 at a time, where p is the number of dimensions of V .
These conditions are quite easy to construct, but that would take me too far
away from my subject.
So if the integral (11) satisfies the conditions (12) in the domain D, the var-
ious values of that integral taken over various closed manifolds of m dimensions
forming part of V will be linear combinations, with integral coefficients, of a
certain number of them, which we could call the periods of the integral (11).
The maximum number of periods is equal to Pm − 1, since, if we consider
Pm closed manifolds of m dimensions there will always be a manifold of m + 1
dimensions which has these Pm manifolds as its boundary, or even some subset
of them. Then there will always be a linear relation with integral coefficients
between the Pm corresponding integrals. Moreover, we can see that there always
exist integrals of the form (11) for which the maximum period is attained. This
means of elucidating the definition of Betti numbers was employed by Betti
himself for the first and the last of these numbers, i.e. for P1 and Pm−1 ; but we
have come to see that it is easy to do the same for the other Betti numbers.

§8. Orientable and non-orientable manifolds

We consider a manifold V defined in the second manner, i.e. in the form of a
chain or network of partial manifolds each of which is defined by relations of
the form (8) and (9).
Let v1 be a partial manifold defined by the conditions

xi = θ(y1 , y2 , . . . , ym ), |yk | < βk

Let v2 be another partial manifold defined by the conditions

xi = θi0 (z1 , z2 , . . . , zm ), |zk | < γk

Suppose that these two manifolds have a common part v 0 forming a con-
nected manifold. I claim that in the interior of this manifold the Jacobian
∂(y1 , y2 , . . . , yn )
∆=
∂(z1 , z2 , . . . , zn )
§8. Orientable and non-orientable manifolds 35

always has the same sign.

In fact, it cannot change sign without vanishing or becoming infinite. We
have
∂(x1 , x2 , . . . , xm ) ∂(x1 , x2 , . . . , xm )
∆= ÷
∂(z1 , z2 , . . . , zm ) ∂(y1 , y2 , . . . , ym )
so that ∆ is itself the quotient of two Jacobians; since these two Jacobians are
essentially finite, ∆ cannot be zero unless we have

∂(x1 , x2 , . . . , xm )
= 0
∂(z1 , z2 , . . . , zm )

and since nothing distinguishes the first m variables x1 , x2 , . . . , xm from the

n − m others, xm+1 , xm+2 , . . . , xn , it will be necessary for the Jacobian of any
m of the x with respect to the z to be zero.
All the Jacobians of the functions θ0 will then vanish simultaneously, contrary
to hypothesis. ∆ then cannot vanish, and we see in exactly the same way that
it cannot become infinite.
Thus ∆ is always of the same sign and we can choose the order of the
variables z in such a way that this sign is positive.
A difficulty may occur in certain cases; suppose that the common part of v1
and v2 , instead of reducing to a single connected manifold v 0 , is composed of
several connected manifolds v 0 , v 00 , v 000 ; in each of them the sign of ∆ remains
constant, but it may change in passing from one to the other. In that case we
say that the manifold V is non-orientable.
We assume that this circumstance does not occur, and consider a sequence
of partial manifolds forming a closed chain

v1 , v2 , ..., vq , v1 .

Let
y1i , y2i , ..., i
ym
be the m variables which play the rôle of y1 , y2 , . . . , ym in relation to vi .
Suppose that v1 and v2 have a common part v10 , v2 and v3 a common part
0 0
v2 , . . . , vq−1 and vq a common part vq−1 and finally vq and v1 a common part
0
vq .
Let ∆i be the Jacobian of

y1i+1 , y2i+1 , ..., i+1

ym

with respect to
y1i , y2i , ..., i
ym .
This determinant will be defined in the interior of vi0 . In the interior of vq0 I also
define the Jacobian ∆q of

y11 , y21 , ..., 1

ym
36 Analysis Situs

with respect to
y1q , y2q , ..., q
ym .
From what we have seen, we know we can always choose the order of variables
so that
∆1 , ∆2 , . . . , ∆q−1
are always positive. On the other hand, though Dq is always of the same sign,
is this sign + or −?
If the sign is −, then I say that the manifold V is non-orientable. I could
also say that the manifolds v1 , v2 , . . . , vq form a non-orientable chain.
Suppose now that we have constructed a certain connected network of partial
manifolds

(4) v1 , v2 , ..., vq

such that each point of V is in the interior (exclusive of the boundary) of one or
more of the manifolds (4). If the determinant ∆ is positive in the common part
of any two of the manifolds (4), I say that the manifold is orientable. If this is
not the case then it is clear that we could always form a non-orientable chain
with some of the manifolds (4), and that the manifold V is non-orientable. I
could also say that the network of manifolds (4) forms an orientable system.
But to justify our definition completely, we have to see that V cannot be
orientable and non-orientable at the same time. It is clear, first of all, that in
the orientable system (4) we cannot find a non-orientable chain.
It remains to show that the system (4) remains orientable when we adjoin
any manifold vq+1 forming part of V .
Let vi0 be the common part of vi and vq+1 ; each point of vq+1 belongs to at
least one of the manifolds vi0 , and, since vq+1 is continued, if I consider two points
M1 and Mk of vq+1 belonging to v10 and vk0 respectively, we can find intermediate
manifolds which I can call (since the numbering remains arbitrary)

v10 , v20 , ..., vk0

and which form a chain.

I can always choose the order of the variables, analogous to the y, defining
the manifold vq+1 in such a way that the determinant analogous to ∆ between
v1 and vq+1 is positive in v10 ; I shall call it ∆1 .
Likewise, we let ∆i be the analogue of ∆ between vi and vq+1 ; ∆i is defined
in the interior of vi0 .
Then let ∆0 be the determinant between v1 and v2 ; it will be defined through-
out the common part of the two manifolds and, in particular, in the common
part of v10 and v20 ; it will be positive because the system (4) is orientable.
Now, in the common part of v10 and v20 , since ∆1 and ∆0 are positive, it
follows that their ratio
∆1
∆2 = 0
∆
§8. Orientable and non-orientable manifolds 37

is positive, and since it always has the same sign, it will be positive for all points
of v20 .
Step by step, we show that ∆3 , . . . , ∆k are likewise positive.
The system thus remains orientable with the adjunction of vq+1 , it also
remains so if the variables y1 , y2 , . . . , ym in the equations for one of the manifolds
vi are replaced by holomorphic functions of new variables y10 , y20 , . . . , ym 0
with
non-vanishing Jacobian (this is necessary in order that a system of values of the
y correspond to a single system of values of the y 0 ). It is necessary, of course,
to choose the order of the new variables y 0 in such a way that this Jacobian is
positive.
The system always remains orientable, we cannot construct a non-orientable
chain, so that a manifold cannot be orientable and non-orientable at the same
time. Q.E.D.
Everyone knows the example of a one-sided (non-orientable) surface obtained
from a rectangle of paper ABCD (where the pairs of opposite sides are AB, CD
and BC, DA) by joining the sides AB, CD so that A coincides with C and B
with D.
Examples of orientable manifolds are easy to construct. Thus in the space
of n dimensions:

10 Every n-dimensional domain is orientable.

20 Every 1-dimensional curve is orientable.
30 Every closed hypersurface of n − 1 dimensions is orientable.

But we can go further.

Consider a manifold V defined in the first manner, i.e. by equations and
inequalities of the form (1). I claim that this will always be orientable.
In fact, let
F1 = F2 = · · · = Fp = 0
be the p equations which, together with some inequalities we shall not write,
define V . Let V be decomposed into a certain number of partial manifolds v,
defined by relations of the form (8) and (9). Let

v1 , v2 , ..., vq , v1

be a certain number of partial manifolds which form a connected chain, i.e. such
that each has a part in common with its successor. I claim that this chain is
always orientable.
In fact, we have assumed above that the Jacobians of the p functions F
relative to any p of the variables x are never simultaneously zero. I may then
assume that the manifolds v are sufficiently small that in the interior of v1 one
of the Jacobians (which I shall call ∆1 ) does not vanish; in the interior of v2
another of the Jacobians (which I shall call ∆2 ) does not vanish, and so on.
If this is not already the case, we can attain it by subdividing each of the
manifolds v into sufficiently small manifolds.
38 Analysis Situs

Suppose, for example, that

∂(F1 , F2 , . . . , Fp )
∆1 =
∂(xα1 , xα2 , . . . , xαp )
∂(F1 , F2 , . . . , Fp )
∆2 =
∂(xβ1 , xβ2 , . . . , xβn )
...............

Since the order of the letters xα1 , xα2 , . . . , xαp remains arbitrary and ∆1 does
not change sign in the interior of v1 , I can always assume that ∆1 is positive in
the interior of v1 . Likewise, ∆2 will be positive in the interior of v2 ; and so on.
Now in the interior of v1 we set
½
F1 = 0, F2 = 0, . . . , Fp = 0
(13)
y11 = x0α1 , y21 = x0α2 , . . . , ym 1
= x0αm

Here m = n − p; the variables x1αi are the m = n − p variables x which

remain when we remove the p variables xαi .
In the interior of v2 we set

F1 = 0, F2 = 0, . . . , Fp = 0,
(14)
y12 = x0β1 , y22 = x0β2 , . . . , ym
2
= x0βm .

The variables x0βi are the m = n − p variables which remain after we remove
the p variables xβi .
And so on.
I assume that the order of the variables x0αi has been chosen in such a way
that we can pass from the normal order of the variables x, i.e.

x1 , x2 , ..., xn

to the order

xα1 , xα2 , ..., xαp , x0α1 , x0α2 , ..., x0αm

by a substitution of the alternating group.

Then, in solving the equations (13) we see that in the interior of v1 the x
are holomorphic functions of

y10 , y20 , ..., 0

ym

and so on. In general, in the interior of vi the x are holomorphic functions of

y1i , y2i , ..., i

ym .

It is then a question of calculating the determinant

∂(y12 , y22 , . . . , ym
2
)
1 1
∂(y1 , y2 , . . . , ym 1 )
§8. Orientable and non-orientable manifolds 39

which we shall abbreviate

∂yi2
∂yi1
in the interior of the common part of v1 and v2 .
To do this, we replace the equations (13) and (14) by the more general
equations
½
0 F1 = λ1 , F2 = λ2 , ..., Fp = λp
(13 )
yi1 = x0αi
and
(140 ) Fk = λk , yi2 = x0βi .
We later put λk = 0.
Solving the equations (130 ) we obtain the x as functions of the λk and the
0
yi holomorphic in the interior of v1 and the Jacobian of
x1 , x2 , ..., xn
with respect to
λ1 , λ2 , ..., λp , yi1 , yi2 , ..., yim
is evidently ∆11 .
Likewise solving the equations (140 ) we obtain the x as functions of the λk
and yi2 holomorphic in the interior of ∆2 , and the Jacobian is ∆12 .
It is then true for points common to v1 and v2 that
∂yi2 2
∂(λ1 , λ2 , . . . , λp , y12 , y22 , . . . , ym ) ∆1
0 = 0 0 0 )
=
∂yi ∂(λ1 , λ2 , . . . , λp , y1 , y2 , . . . , ym ∆2
and it only remains to make λk = 0 in the expressions for the x.
We similarly find that in the part common to vk and vh+1
∂yih+1 ∆h
=
∂yih ∆h+1
and in the part common to vp and v1
∂yi1 ∆q
q = .
∂yi ∆1
We have already seen that we can always assume that ∆i is positive in the
interior of vi . All these quotients are then positive and the chain is orientable.
Q.E.D.
Thus all varieties which satisfy the first definition are orientable, and, since
I cited an example of a non-orientable manifold satisfying the second definition
we conclude that there are manifolds satisfying the second definition which do
not satisfy the first. This is what I claimed originally.
Any non-orientable manifold is, by definition, opposite to itself.
40 Analysis Situs

§9. Intersection of two manifolds6

Let V and V 0 be two manifolds defined in the second manner, one having p
dimensions, the other n − p, and let M0 and M00 be two points belonging to V
and V 0 respectively with respective coordinates

x01 , x02 , ..., x0n

and
x00
1 , x00
2 , ..., x00
n.

Around the point M0 we construct a partial manifold v on V , analogous to

those envisioned in paragraph 3, in such a way that for a point x1 , x2 , . . . , xn of
v we have
xi = θi (y1 , y2 , . . . , yp ) (i = 1, 2, . . . , n).
Likewise, around M00 we construct a manifold v 0 on V 0 so that for a point
x01 , x02 , . . . , x0n of v 0 we have

x0i = θi0 (y10 , y20 , . . . , yk−p

0
) (i = 1, 2, . . . , n).

Consider the determinant

¯ ∂x1 ∂x2 ∂xn ¯
¯ ∂y ... ¯
¯ 1 ∂y1 ∂y1 ¯
¯ ¯
¯ ∂x ¯
¯ 1 ∂x2
. . . ∂x n ¯
¯ ∂y2 ∂y2 ∂y2 ¯
¯ ... ... ... ... ¯
¯ ¯
¯ ∂x1 ∂x2
. . . ∂x n ¯
¯ ∂yp ∂yp ∂yp ¯
¯ ¯
¯ ¯
¯ ∂x0 ∂x02 ∂x0
¯
¯ 1
. . . ∂yn0 ¯
¯ ∂y10 ∂y10 ¯
¯ ... ... ... ...
1
¯
¯ ¯
¯ ∂x01 ∂x02 ∂x0 ¯
¯ ∂y0 0
∂yn−p . . . ∂y0 n ¯
n−p n−p

0
This is a function of y1 , y2 , . . . , yp , y10 , y20 , . . . , yn−p ; thus it depends on the posi-
tion of the points M and M with coordinates x1 , x2 , . . . , xn and x01 , x02 , . . . , x0n
0

on the manifolds v and v 0 . I therefore call it f (M, M 0 ).

I can change the variables by replacing y1 , y2 , . . . , yp by holomorphic func-
tions of p variables z1 , z2 , . . . , zp chosen in such a way that a system of values of
the y corresponds to a single system of values of the z. For this it is necessary
that the Jacobian of the y with respect to the z never vanishes, and I can al-
ways assume that the z are arranged in an order which makes the determinant
positive.
The function f (M, M 0 ) is then multiplied by this Jacobian, and consequently
retains its sign.
6 The Supplement to Analysis situs takes this up again.
§9. Intersection of two manifolds 41

It is the same when we make an analogous change of variables on the manifold

v0 .
Consider now another manifold v1 , analogous to v, constructed on V , and a
manifold v10 , analogous to v 0 , constructed on V 0 .
Let M1 be a point of v1 and M10 a point of v10 . We can construct a function
analogous to f (M, M 0 ) which I shall call f1 (M1 , M10 ).
Suppose now that the manifolds v1 and v (likewise the manifolds v10 and v 0 )
have a common part and that the points M and M 0 are identified, likewise the
points M 0 and M10 .
We compare the two functions
f (M, M 0 ), f1 (M, M 0 ).
Suppose that the two manifolds V and V 0 are orientable. We may then
assume (by suitable choice of the order of the variables y relative to v1 ) that
the determinant analogous to that we called ∆ in §8, relative to v and v1 , is
positive on the part common to v and v1 ; likewise, the determinant analogous
to ∆ relative to v 0 and v10 will be equally positive.
Then f and f1 have the same sign.
Then let S(M, M 0 ) be a function which is equal to +1, −1 or 0 according
as f (M, M 0 ) is positive, negative or zero. This function is well-defined, it does
not depend on the position of the points M and M 0 on V and V 0 ; nor does it
depend on the manner in which the manifolds v and v 0 have been constructed
around M and M 0 .
This is no longer true if one of the manifolds V and V 0 is non-orientable.
Suppose in particular that the points M and M 0 are the same, so that the
point M is a point of intersection of V and V 0 ; consideration of the function
S(M, M 0 ) is then of great interest.
Imagine that we construct the function S(M, M ) for all points of intersection
M of V and V 0 , and take the sum of all the functions S so obtained; I shall
denote this sum by N (V, V 0 ).
This definition of N (V, V 0 ) is applicable when the manifolds V and V 0 have
been defined as in §3. But we can simplify it when V and V 0 have been defined
as in §1.
Let
Fα = 0, ϕβ > 0, (α = 1, 2, . . . , p; β = 1, 2, . . . , q)
be the conditions which define V and let
Fγ0 = 0, ϕ0δ > 0 (γ = 1, 2, . . . , n − p; δ = 1, 2, . . . , q 0 )
be those which define V 0 . V has n − p dimensions and V 0 has p.
Consider a point M of V with coordinates x1 , x2 , . . . , xn and a point M 0 of
V with coordinates x01 , x02 , . . . , x0n and form the determinant
0
¯ ¯
¯ ∂F1 ∂F2 ∂F10 ∂F20
0
∂Fn−p ¯
¯ ∂x1 ∂x1 . . . ∂F p
0 0 . . . 0 ¯
¯ ∂x1 ∂x1 ∂x1 ∂x1 ¯
¯ ... ... ... ... ... ... ... ... ¯
¯ ¯
¯ ∂F1 ∂F2 ∂Fp ∂F10 ∂F20
0
∂Fn−p ¯
¯ ∂xn ∂xn . . . ∂xn ∂x0 ∂x0 . . . ∂x0 ¯
n n n
Other documents randomly have
different content
Who said twas only frogs.

Be that as each opined, ’tis sure

With Demus soon it turned
To ague, and the only cure
For flesh which froze or burned,
The doctor ordered, was to drain
The hollow in the rear
Where Demus lived; for while in vain
He followed his career
Of human welfare, there had lain
The most neglected near.
’Twas remedied and ne’er again
Did Nicodemus hear
The voice which had become so famous
For back-door croaks and frogs
Call
——“Nicodemus! Nic-o-de-mus!
Nic-o-de-mus-s Bog-g-s!”
SA C R E D
WHAT IS FAITH?
Faith is no weakling, howsoe’er
It needeth courage for its task,
But strength whose confidence to dare
Is that which humbles it to ask
A higher help, a higher word
To lift it, bid it trust and try,
Assured its selfless prayer is heard,
Its task beneath a Master’s eye.

Faith is the reasoning of heart

Toward the Heart-of-hearts which beats
In unison with every part
Of all it quickens and completes;
And with a sense of love and plan
Sees only good from truth and right,
Wrong as the only ill which can
Defeat design and quench the light.

Faith is the fortifying gate

Which walls us in, our terrors out,
Through which we fare to conquer fate
Or flee for refuge from our doubt;
Faith blows the trumpet, mans the tower,
Inspires hope, believes in Heaven
And trusts the overruling Power
To care for what its will hath given.

Faith is the burden-bearer’s stay,

The footsore pilgrim’s trusty staff,
The victor’s martial panoply,
The martyr’s noblest epitaph.
Faith is the vision’s inner eye
Whose pupil is the seeing soul,
Its iris the reflected sky,
Its long perspective Spirit’s goal.
A FO RGIVENESS
A pilgrim long devout arrived at last
Before the Gate of Paradise, and cast
His staff aside triumphantly to press
Within the dreamed-of goal. But strange to say,
It did not open to his eagerness
As knocking he solicited the way.

“Nay,” said the Guardian Angel of the Gate,

“The proof of thy assurance I await,
The sesame and heavenliest word
That passes here! Three trials shalt thou have,
And if thou hast not found it by the third
No privilege to enter canst thou crave.”

So sure the Pilgrim was the truest right

Must be the one of evangelic might
He quickly answered “love!”
The Angel’s wing
Drooped o’er his countenance as he replied,
“Nay, such a plea might any sinner bring
Like any saint whose zeal is undenied.

“Canst thou not to the name come closer yet

Of Goodness’ greatest key?”
The Pilgrim let
His thoughts go outward in a second quest
And slowly made response, “Why, then, ’tis grace,
The covenant and seal of all the rest,
The chain whose lock is Love.”
The Angel’s face
Was still compassionate as he withheld
The entrance, and his pity would have spelled
The password in his eyes as he again
Made answer, “Grace is truly all our hope
In promise and fulfilment, but ’tis when
We lay it to our hearts the Gate we ope
We lay it to our hearts the Gate we ope
And our admission most divinely plead;
For none can think the word but feels its need
And healing touch.”
The Pilgrim’s brow grew sad,
But as he pondered to his knees he fell
And rose as oft before in wonder glad—
“Forgiveness!”
The Angel answered, “Well!”
And stood aside to let him pass.
TH E GO OD SAMARITAN
The Good Samaritan was he
Who had compassion not alone
Humanely but divinely. We
Must look beyond the Healer—see
The Sympathizing Savior—be
Forgiven, lifted up and shown
The heart of Love and in our own
Begin to feel the sympathy
Which from His humanness had grown
To deeds of such divinity.

How little ’tis to minister

To one poor soul unless we feel
The touching brotherhood of care,
The sense how easy ’tis to err,
To fall, to need another’s prayer,
Another’s help! But when we kneel
Our fellowfeeling must be real
Enough that we can rise and share
The burden of our own appeal
And help our brother’s cross to bear.

He is the Good Samaritan

Who loves enough to never wrong,
To ever right a brother man—
To bind his wounds and shape the plan
Of life benignly so he can
His neighbor also cheer along.
Blest be the mercifully strong!
Blest be the human-hearted man
Who never quenched a living song!
For he is God’s Samaritan.
 SHEP HER D OF ISRAEL

Shepherd of Israel, hear

The calling of thy flock,
And when we seek do thou be near
To lead us to the Rock
Where full and sheltered we
At noonday may repose
Or find at night security
From all our lurking foes!

Help us to trust thy care

Through green or barren ways
And voice our doubts and fears in prayer,
Our blessedness in praise!
If thorns beset our path,
To feel Thou leadest us
Is sweet assurance goodness hath
A loving purpose thus.

Guide us by living streams

That rise in mountain height
And up where wisdom’s heavenly beams
Our spirits bathe in light!
Lead us to ranges high,
To visions rich and broad,
To pinnacles that touch the sky
And help us know Thee, God!
TH E LADDER OF CLOUD
There’s a beautiful ladder of fine-spun cloud
That stretches from earth to sky
And up and down it the angels crowd
With calling and soft reply:—

Amrael

Children of men, who only by sight

Know that the stars exist,
There was one that shone o’er the world last night
Through an aureole of mist.

Mishael

They only saw it who had kept

The vigil of the seers
With inner sense; but ye who slept
Knew not the sign of the years.

Uriel

The spirit of life became a star

And we the herald-host;
And we sang as the Wise Men gazed afar
And the Shepherds Heavenmost;

Host

Joy to the world! For lo, is born

The Gift-Child! Echo on
And on forever song of morn,
Yet trembling into dawn!

Refrain

Joy to the pure in heart! For thou

Alone dost know the worth
 o e dost o t e o t
And meaning of the Gift, who bow
Before the Virgin-birth.

Chorus

All hail Madonna’s Gift

That shall the earth to Heaven uplift!
All hail! Rejoice!

What softening of angel-voice

And light and listening sense
Fell hush-like on the last “Rejoice,
Madonna-reverence!”

The pearly wings the host enshroud,

The voices fade away,
And the beautiful ladder of fine-spun cloud
Becomes the Gate of the Day.
THE R ISEN C HRIST MEANS VICTORY

Go forth and hail the Conqueror

With flowers and sacred psalms!
The triumph we observe is more
Than that of martial palms;
For lo! there cometh from the tomb
The Lord of life and life-to-be,
Around whose feet the lilies bloom;
The risen Christ means victory.

Go forth and on His living brow

Entwine a laurel-wreath;
For never was so great as now
The glory of His death!
The Cross and Sepulchre had been
The world’s most damning tragedy
But for the conquered curse of sin;
The risen Christ means victory.

Go forth with precious ointment of

Affection to thy dead,
With Easter’s glad, believing love
That He Who for us bled,
Who slept and rose again, is strong
To roll corruption’s stone away.
And loose the Resurrection Song;
The risen Christ means victory!
 TH E EVER LASTING AR MS

When to our life dark shadows come,

Stern crosses, sacrificial cares
And other fancied temporal harms,
There is eternal refuge from
Our terrifying doubts and fears
Within the Everlasting Arms.

When o’er our souls temptations sweep

And goodness loses half its grace
As sin pursues us with its charms,
There is no refuge left to keep
But the eternal hiding-place
Within the Everlasting Arms.

When through the valley dark and drear

We walk or see another sink
And death o’ercomes us with alarms,
Be then, Eternal Refuge, near
To hold us up upon the brink
Within the Everlasting Arms!
HE GIV ETH HIS BELOVED SL EEP

The task is done, the sun is set,

The evening shadows fall apace,
The course is run, and tarries yet
The glory only of the race;
But ere the guerdon of the toil
The fleeting soul shall rise to reap,
God maketh it to rest awhile—
He giveth his beloved sleep.

What though the eyes are closed in death,

The tired hands are folded now?
Life shall arise, saith living faith.
And ministry diviner grow.
’Tis but the hush before the day:
The Father bids his angels keep
The treasure that we lay away—
He giveth his beloved sleep.

But not, oh not forever thus

Doth death enshroud our silent ones—
We know not what transfigures us,
What miracle of quickening suns—
But we await their healing wings,
Their living flash, seraphic sweep,
The glory of the King of Kings
Who giveth his beloved sleep.
 TH E GLO RY DWEL L S

Oh, the glory that we dream of

Trembling over Bethlehem!
Magi following the beam of
Starry prophecy to them!
Shepherds startled by the gleam of
Heavenly light and angel-hymn!

Time hath made the vision holy,

But I know that glory dwells
Not in manger-village solely,
Nor in dream that prophet tells,
But wherever there’s a lowly
Child-heart, there the glory swells.

Pride of earth and pomp of power

Dazzle with their tinsel show;
But compared to goodness’ dower
They’re as only glint to glow.
Pride is merely for an hour,
Goodness doth to glory grow.
 TH E L IGHT OF L IFE

O Light of Life, shine thou

Into my soul as doth the Sun of Day
Into the world for seeing with mine eyes!
Reveal the good and evil—teach me how
To stumble not but walk the Living Way
That fills earth with the glory of the skies!

Let there be spirit-quickenings

That thrill the being to responsiveness
Lest vision be but human, uninspired!
Ah, make it throb until from vision springs
Anointed nature to in life express
The Grace which makes the Heavenly desired!
DESIGN
The universe of rolling spheres
Is not for Deity’s display
But for a purpose which appears
In its supernal harmony.

Its mass that in momentum sweeps,

Its energy of elements,
The order which its system keeps
Are aspects of omnipotence;

And power working such design

Is proof of Presence everywhere
Intelligent, supreme, divine,
Both in creatorship and care.

For in His watchcare of the worlds

He-Over-All doth manifest
A greater power than that which whirls
Them on their way at its behest,

A greater purpose than to span

The Heavens by His glory lit;
For ’tis the more eternal plan
Of making all creation fit

For fellowship with Nature’s God

In higher terms of wisdom, truth
And love by perfect will endowed,
Whereof the worlds are but the proof.

Thou Supersoul, who Spirit art

And rulest star-host, wave and wind,
Teach us Thy majesty to heart
And feel in music perfect Mind!
SONG
 GO L DEN HOP E

There is nothing in the world so sweet

As the hope which never, never dies,
That sometime, somewhere we shall meet
In gladder love beyond the skies—
Oh, beyond the skies so golden,
With the hope of Heaven olden;
For there’s nothing in all the world so sweet
As the olden, golden hope again to meet!

There is nothing in all the world so fleet

As the hope that ever, ever flies
Swift onward, upward to the seat
Of perfect love beyond the skies—
Oh, beyond the skies so glowing,
With the hope of Heaven growing;
For there’s nothing in all the world so sweet
As the glowing, growing hope again to meet!

There is nothing in all the world so great

As hope that bids us, helps us rise
With more responsive hands and feet,
With gladder tongues and clearer eyes—
Oh, upon the skies so golden,
With the hope of Heaven olden;
For there’s nothing in all the world so sweet
As the olden, golden hope again to meet!
TH E COMING CR OWNING
When the chariots of glory
Come flashing from the east
On the day of Advent-story,
The crowning of the Christ;
When the clouds are seraph-mounted
And radiant of wing
With angel-hosts uncounted,
And the skies with rapture ring—
My soul, wilt thou undaunted
Meet the coming of the King?

When earth the blessed vision

With lifted eyes beholds
And feels the swift transition
Of glory that enfolds;
When from the skies descending
The hosts of Heaven bring
The Kingdom never-ending
Of which all peoples sing—
O Spirit, wilt thou blending
Hail the coming of the King?

When thrones are set for mercy

And love to minister
To the naked, sick and thirsty
And all who faint or err;
When the Lord of glory reigneth
And choired censers swing
With the praises God ordaineth
As Heavens their banners fling—
O Soul, a crown that gaineth,
Crown and enthrone the King!
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebookgate.com