You are on page 1of 24

Ontology and Mathematical Practicet

JESSICA CARTER*

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


1. Introduction
In a recent book, David Corfield urges philosophers of mathematics to take
the approach of those he denotes as
the practice-oriented philosophers, or what I am calling philosophers of real
mathematics. Continuing Lakatos's approach, researchers here believe that a
philosophy of mathematics should concern itself with what leading mathemati-
cians of their day have achieved, how their styles of reasoning evolve, how they
justify the course along which they steer their programmes, what constitute
obstacles to these programmes, how they come to view a domain as worthy of
study and how their ideas shape and are shaped by the concerns of physicists
and other scientists. (Corfield [2003], p. 10)
This is the approach taken in the present paper, as its aim is to propose
an answer to the question of the nature of mathematical objects, based on
a case study in part of modern mathematics.
The paradigmatic example of using the history of mathematics to draw
conclusions about the nature of mathematics was set by Lakatos. In his
Proofs and Refutations he presented a study of the history of Euler's theo-
rem, challenging the formalist conception of mathematics, and proposing a
fallibilist view of mathematics. Lakatos's ideas have inspired many philoso-
phers, and his work has been discussed recently by Corfield [2003], Larvor
[2001] and Leng [2002a]. Larvor acknowledges the importance of Lakatos,
but points out that 'his significance lies, rather, in having turned philo-
sophical attention to what one may call the "inner life" of mathematics'
(Larvor [2002], p. 213).
Thus, there is an increasing number of scholars who hold that philoso-
phers of mathematics should study practice.1 Nevertheless, when it comes

T Thanks to S. Andur Pedersen and H. J. Munkholm, who in various ways have con-
tributed to the material presented here. I also wish to thank the two anonymous referees
for their helpful criticism of earlier versions of this paper.
* Department of Curriculum Studies, Danish University of Education, Copenhagen,
Denmark, jeca@dpu.dk
1
The recent volume The Growth of Mathematical Knowledge edited by Grosholz and
Breger is another outcome of this philosophy at work.

© 2004 PHILOSOPHIA MATHEMATICA (3) Vol. 12, pp. 244-267.


ONTOLOGY AND MATHEMATICAL PRACTICE 245

to the issue of which questions can be answered through such studies, there
is no general consensus. Larvor denies that it is possible to draw conclusions
about the ontology of mathematics:
Whether we adopt fictionalism; out of the activities of the mathematicians;
or whether we think of progress as ever-closer approximation to a pre-existing
Platonic reality, makes no difference to our study of the inner logic of math-
ematical development. The dialectical stories turn out the same regardless of
any ontological commitment. (Larvor [2001], p. 218)
In opposition to this statement, it may be remarked that at the beginning

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


of the last century the mathematician H. Weyl based his mathematics on
convictions about the ontology of mathematics. In contrast, Leng [2002a],
though cautious about the kind of conclusions that may be drawn from
practice, does attempt to identify some features of the nature of math-
ematical objects from a case study concerning C*-algebra. As we shall
see later, my conclusions differ somewhat from those of Leng. Finally,
there is M. Muntersbjorn [2003] who does not hesitate to draw conclusions
about the emergence of mathematical objects from a case study concerning
seventeenth-century analysis.
I am sympathetic to the view that it may not be possible to draw defini-
tive conclusions about the nature of mathematical objects from mathemati-
cal practice. My starting point was to study mathematical practice without
any preconceptions to see whether it was possible to detect an answer to
the question of the nature of the mathematical objects. Thus my question
became: What happens when new objects are introduced into mathemat-
ics? The subject chosen for study needed be part of modern mathematics.
I chose ivT-theory, which is used in many different branches of mathematics
and is today regarded as a very important field.

2. Case study in if-theory


i<f-theory started out as a discipline in topology through the work of Sir
Michael Atiyah in the late 1950s. Later lf-theory was introduced into al-
gebra, into C* -algebra, and successfully used in various branches of mathe-
matics. In i^-theory, a series of groups, the so-called Jf-groups, is associated
to an object, for example, a topological space, a manifold or a C*-algebra.
/f-theory has been of great utility in many areas of mathematics. With its
help, a number of hitherto unresolved problems have been resolved. A few
examples will suffice to illustrate this: In 1961, J. Adams managed to find
an expression for the number2 of linearly independent vector fields on the
sphere Sn in E n + 1 . It had been known since the beginning of the twentieth
century that this number of vector fields could be defined on the sphere,
2
The number is p(n + 1) - 1, where p(m) is defined as: If m = {21 + 1) • 2C • 16d, where
I, c, d e Z and 0 < c < 3, 0 < I, and 0 < d, then p{m) = 2l + 8d.
246 CARTER

but it had long been an open and famous problem to prove that more could
not be defined. In 1961, Milnor disproved an assumption dating back to
Poincare, called the Hauptvermutung, that two polyhedral decompositions
of a space always have a common subdivision. Also, if-theory was used by
Atiyah and Hirzebruch in proving that the complex projective space QPn
cannot be embedded in M.N for N =4n- 2a (n), where a(n) is the number
of terms in dyadic expansion of n. In both-high-dimensional geometry (the
study of manifolds of dimension larger than 5) and in C*-algebra, if-theory
is an important tool in classifying manifolds and algebras, respectively.

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


Atiyah's work was based on the work of Alexander Grothendieck, who
introduced the first .ff-group. Because of their work on X-theory, Grothen-
dieck and Atiyah won the Fields medal in 1966.3 Our main interest here
will be to determine why Grothendieck introduced the first if-group. The
history of the introduction of Jf-theory is presented in more detail in Carter
[2002].
There is a previous paper by J.-P. Marquis [1997] which, using if-theory
as an example, draws conclusions about the nature of mathematics. Mar-
quis claims that it is not only in science that tools are used to determine
the properties of objects, if-theory, for example, can be regarded as a tool
or a machine that mathematicians use to determine the properties of math-
ematical objects. I agree that K"-theory can be viewed as a tool; compare
the examples given above. One of the conclusions that Marquis draws from
this observation is that an ontological distinction can be drawn between
the objects of the mathematical universe:
If we admit that there are instruments and machines within this universe, then
we grant the fact that there are mathematical 'natural kinds' and mathematical
'artefacts'. (Marquis [1997], p. 270)
This distinction, Marquis argues, is problematic to Platonists. I do not
believe that the distinction affects the point that I intend to make, namely
that mathematical objects are introduced by mathematicians. It may well
be that some objects/entities seem more natural, for example, because they
arise as abstractions of objects that are studied in the natural sciences,
whereas others may seem more abstract and artificial. It should also be
emphasized that one has to distinguish between objects and theories. K-
theory, as mentioned, began with the introduction of an object, the K-
group, and was only later generalized into a theory. One point that Marquis
makes is that if-theory was invented, but I hold that this holds for all
mathematical objects. In my view, the interesting question, when drawing
such a distinction, is this: If the tools or machines of mathematics seem
artificial, then how are mathematicians able to create them and where do

3
In 1978 Quillen was also awarded a Fields Medal for his development of algebraic
if-theory.
ONTOLOGY AND MATHEMATICAL PRACTICE 247

they get their ideas from?4


The If-group was introduced by Grothendieck in an attempt to general-
ize the so-called theorem of Riemann-Roch. This theorem was first posed
by Riemann in the 1850s (Riemann [1892]). Inspired by the topic of Abelian
integrals, he introduced a kind of surface that was later named a Riemann
surface and a method of cutting up the surface such that it becomes simply
connected. A surface is called simply connected if any closed curve on the
surface can be continuously shrunk into a point while still being on the
surface. He could then establish the existence of complex functions defined

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


on such a surface depending on the number of poles of the function and the
number of cuts on the surface. Roch [1865]] completed the theorem when
he was able to determine the number of constants on which such a function
could depend.
Riemann's methods were intuitive and not very rigorous. Later Hermann
Weyl wrote a book on the subject Die Idee der Riemannschen Flache (Weyl
[1913]), where it was given a rigorous treatment.
As a Riemann surface is a 1-dimensional complex surface, i.e., a curve,
Riemann and Roch's theorem is a theorem for curves. Mathematicians
asked whether it was possible to find a generalization for surfaces of arbi-
trary dimensions. This generalization was not possible until 1956, when
Hirzebruch published his version of the theorem (Hirzebruch [1966]). To be
able to prove this theorem many new notions/objects had to be introduced.
One of these was sheaves, introduced by the French mathematician J. Leray
(Leray [1946]). As mentioned, the theorem of Riemann-Roch establishes the
existence of complex functions on a surface in terms of geometric invariants;
in Riemann's case these were the number of cuts to make the surface simply
connected. Another problem that had to be dealt with when generalizing
the theorem was the definition of such invariants for surfaces of dimension
greater than 1. This led to the definition of the arithmetical genus of a
surface.

2.1 Grothendieck's definition of the i^-group


Early in his career, Grothendieck was influenced by the Bourbaki group.
After graduating from the University of Montpellier in 1948, Grothendieck
spent some time in Paris, where he participated in Henri Cartan's famous
1948-1949 seminars on sheaves. At that time Grothedieck was more in-
4
There is another interesting aspect to this question: Often when mathematicians talk
about their work, they use metaphors; for example, Marquis quotes A. Connes saying
that mathematicians create tools. The question is whether these descriptions are more
than mere metaphors. Marquis argues rather convincingly that K"-theory can indeed be
classified as a tool. But is this always the case? Turning the question upside down, it
may be asked whether the reason that mathematical methods resemble scientific methods
is because it is mathematicians who create mathematics, and in doing so they may be
inspired by the methods that are used in science.
248 CARTER

terested in functional analysis. So, on the advice of Cartan, he went to


Nancy in October 1949. Here were J. Delsarte and J. Dieudonne, who
with Cartan were the founding fathers of Bourbaki. According to Cartier
[2000], the close connection with Bourbaki (Grothendieck also became a
member) inspired Grothendieck's sense of generality in mathematics. Af-
ter completing groundbreaking work in functional analysis, Grothendieck
became interested in algebraic geometry with J.-P. Serre as his most im-
portant collaborator. Serre was one of Cartan's students and early became
involved in the theory of sheaves. Serre's fundamental article 'Faisceaux

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


algebriques coherents', was published in 1955 (Serre [1955]). This article
gives the foundation for the theory of sheaves in algebraic geometry, and
describes the theory of coherent sheaves defined over an algebraic variety
defined over an arbitrary field. Grothendieck had been in contact with Serre
for some years, and they had exchanged letters since 1954. Grothendieck
quickly made progress in this area, and it is at the start of this work that
he formulated the generalization of the theorem of Riemann-Roch, which
is our main interest.
Another of the characteristics of Grothendieck's style was to replace theo-
rems about objects, so-called absolute theorems, with theorems about func-
tions, so-called relative theorems. This trend was inspired by the notions
from category theory which emerged around that time. Mathematicians
were beginning to study relations between objects instead of the objects
themselves, in particular what could be said about the objects when there
is a morphism (this is the word used in category theory for functions) be-
tween them. Category theory arose because mathematicians became aware
of the structural similarities between the various branches of mathematics.
In all branches, mathematicians study certain kind of objects and maps be-
tween these objects, for example, topological spaces with continuous maps
or groups with homomorphisms. In category theory, the structural similar-
ities between all kinds of objects with maps defined on them are studied.
Therefore, a category is defined as a class (or a universe) of objects with
some morphisms defined between these objects, which are subject to certain
conditions. Then the usual notions of, for example, a kernel of a morphism
or a product is defined through these morphisms. Once the notion of a cate-
gory is defined, one defines maps between categories, the so-called functors.
Finally maps between functors, natural transformations, can be defined.
S. Mac Lane and S. Eilenberg were the mathematicians who initiated the
development of category theory. The first notion that was introduced was
that of a natural isomorphism (see Eilenberg and Mac Lane [1942]) and
later (in Eilenberg and Mac Lane [1945]) categories were defined for the
first time.
Thus, items that inspired Grothendieck's formulation of the Riemann-
Roch theorem were:
ONTOLOGY AND MATHEMATICAL PRACTICE 249

• Hirzebruch's theorem of Riemann-Roch.


• The rise of category theory proposing a study of so-called relative the-
orems, i.e., theorems about morphisms instead of absolute theorems,
i.e., theorems about objects.
• Grothendieck's taste for generality, which partly was inspired by his
close connection with the Bourbaki-group.
• Serre's work on algebraic geometry.
Hirzebruch's work was in algebraic geometry using topological methods.
The basic objects that he worked on were algebraic manifolds, and the struc-

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


tures defined over these are vector bundles. The corresponding notions in
algebraic geometry are those of algebraic varieties and sheaves. Another
difference is that Hirzebruch uses so-called analytical methods, which es-
sentially means that the objects are defined over the complex numbers,
whereas in algebraic geometry the objects are defined over an arbitrary
field.5 So, being a relative version of Hirzebruch's theorem, Grothendieck's
version should be formulated when there is a proper morphism / : X -> Y
between two algebraic varieties, and when there is defined a coherent sheaf6
over X.
Therefore the problem that Grothendieck set himself was (loosely): Find
a formulation of Riemann-Roch, when there is given a morphism f : X —> Y
between the algebraic varieties X and Y, and when there are defined coher-
ent sheaves over X. This formula should reduce to Hirzebruch's theorem
of Riemann-Roch.
The theorem was completed in a series of letters between Grothendieck
and Serre, and presented at a seminar held at Princeton in the fall of 1957.
From the notes of this seminar, A. Borel and J.-P. Serre wrote the article
'Le Theoreme de Riemann-Roch' (Borel and Serre [1958]). Grothendieck
has also written a paper on his theorem of Riemann-Roch, printed in the
Seminaire de Geometrie Algebrique (Grothendieck [1971]). As Borel and
Serre's article is the first presentation of the theorem, we will follow this.
In the above description, I mentioned algebraic varieties and sheaves,
which were the main objects of algebraic geometry when Grothendieck de-
fined the if-group. The definitions of these objects can be found in the
appendix.

5
This often makes proofs shorter and more elegant, and not more complicated as one
might expect!
6
Hirzebruch defines cohomology groups with coefficients in sheaves of germs of local
sections of vector-bundles. More generally (in algebraic geometry), these sheaves would
be locally free sheaves. But supposing that there is a morphism / : X —> Y and that
there are defined locally free sheaves over X, then the corresponding sheaves over Y will
not generally be locally free sheaves. Instead Grothendieck chose to work with so-called
coherent sheaves. For coherent sheaves, Grothendieck had already proved (Grothendieck
[1956/57]) that, if / : X -> Y is a proper morphism, then the sheaves on Y corresponding
to coherent sheaves on X will also be coherent sheaves.
250 CARTER

2.2 Definition of the if-group


Grothendieck wanted to find a formulation of the theorem of Riemann-Roch
when given a proper morphism / : X —> Y between two algebraic varieties
(non-singular, irreducible and quasi-projective7) X and Y, and when there
is defined a coherent sheaf over X. A morphism / : X -¥ Y is said to be
proper if its graph Gf is closed in P x Y where P is a projective space
containing X. That the morphism / is proper has consequences both for
the formulation and the proof of the theorem.

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


The AT-group is introduced when generalizing the Euler-Poincare char-
acteristic appearing on the left-hand side of Hirzebruch's formula:8

For a relative version (focusing on morphisms), what Grothendieck needed


to do was to define a homomorphism ft which given a sheaf over X should
produce a sheaf over Y. Also, being a generalization of Hirzebruch's for-
mula, the expression of the homomorphism should somehow reduce to the
Euler-Poincare characteristic.
The first step, then, was to define a sheaf over Y, when given a sheaf
T over X. This was not a problem, as a method had already been found
by Leray in 1946 when he defined sheaves. Given a sheaf T over X, it
is possible to define the so-called direct image of T, /*(?), which is a
sheaf defined over Y. The procedure is described in Seminaire H. Cartan
(Grothendieck [1956/57], p. 5). These sheaves generally do not reduce to
the expression 5Zj(—I)1 dimH'(X, !F), but when Y is a point, p, such that
we have / : X -»• p, then / . ( ^ = H°(X,J).
However, given the direct image f*(J-) over Y it is possible to derive
a whole sequence of sheaves, Rqf*(J-), over Y, q being a natural number.
These sheaves, Rqf*(?), are called the higher direct images of / , and f*{!F)
is denoted i?°/*(J).
The main point now is that when J- is a coherent sheaf and f : X —¥ Y
is a proper morphism then:
1. i?«/*(-7r) is zero when dim(X) < q.
2. When Y is reduced to a point, such that we have f : X -> p, then the
higher direct images 1??/*(.?) correspond exactly to the cohomology-

7
A variety V is irreducible if, whenever there is an equation V = U U W, then either
V = U or V = W. A variety is said to be quasi-projective if it is isomorphic to a locally
closed subvariety of a projective space. For varieties defined over the complex numbers
a non-singular variety is a complex manifold.
8
Hirzebruch showed that the Euler-Poincar6 characteristic x(^i W)i where V is an
algebraic manifold and W is a vector-bundle over V, could be expressed as a certain
polynomial in the Chern classes of V and a cohomology class determined by W (Hirze-
bruch [1966], p. 155).
ONTOLOGY AND MATHEMATICAL PRACTICE 251

That Rqft (T) is zero when dim(X) < q entails that there are only a finite
number of these sheaves that are different from zero and, since the higher
direct images reduce to the cohomology groups in Hirzebruch's formula, the
natural choice for Grothendieck would be to form the element

i=0

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


as the image of T? This expression, however, is not meaningful over Y
(a linear combination of sheaves is not a sheaf!). To be able to form this
element and to make sure that the map from sheaves over X to sheaves
over Y is in fact a homomorphism, Grothendieck proceeded to define the
if-group as a quotient:
K(X) -
K{x)
~
where E{X) is the free abelian group generated by isomorphism classes of
coherent sheaves over X. Elements in E(X) thus have the form ^ m* * Ti
for m, € Z (only a finite number different from 0) and Ti coherent sheaves
over X. Thus, over E{X) it makes sense to form the alternating sum
above. But defining the function f\ over E(X) will not make it into a
homomorphism, therefore the quotient is necessary. Q(X), then, is the
subgroup generated by expressions of the type

T-T'-T"

whenever there is a short exact sequence10

0-> T'^T^T"^^ (1)

of coherent sheaves over X. This means that, whenever there is such a


short exact sequence (1) of coherent sheaves over X, then the equation

holds in K(X).
9
Comparing this expression with the Euler-Poincar6 characteristic in Hirzebruch's for-
mula, it will be noted that the dimension of the cohomology groups is missing from
the former expression. But when considering Grothendieck's generalization as a whole,
J,(ch{x).T(X)) = ch(f,(x)).T(Y), the right-hand side will reduce to the Euler-Poincare
characteristic in Hirzebruch's formula. Here ch denotes the Chern class and T(X) the
Todd polynomial.
10
A sequence . . . —> Cj_i -^> C; ^±> Ci+i... is exact at Ct if im(di) =ker(d< +1 ). If
the sequence is exact at Cj for all i, then the sequence is said to be exact.
252 . CARTER

Grothendieck then defined the map

/, : K(X) -> K(Y)

as taking an element T from K(X) to the element

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


t=0

defined in K(Y).
Because of the way in which the higher direct images are obtained from
the direct images, and because of the properties of exact sequences (see
details in Carter [2002], pp. 70-73), the function f\ will now become a
homomorphism.
This ends the description of how Grothendieck introduced the if-group.
We have seen that, on the basis of developments in various branches of
mathematics, Grothendieck formulated a problem, namely to generalize
the theorem of Hirzebruch-Riemann-Roch. The generalization was inspired
by Grothendieck's taste for generality, his use of algebraic methods, and
the rise of category theory. To solve this problem he introduced a new
object, the Jf-group. As we have seen, the if-group was introduced to
be able to formulate the expression of the function f\ such that it reduces
to part of Hirzebruch-Riemann-Roch.11 It is reasonable to state that in
this case the .ftT-group grows out of mathematical practice. It is not only
the if-group which exemplifies this; there are more examples from the
case study. These are, for example, Riemann's introduction of his space,
the introduction of the arithmetical genus, and the sheaf. Riemann spaces
were introduced to be able to work with Abelian functions. These functions
had the property that, when considered as functions on the complex plane,
they were multiple-valued. In Riemann's time, mathematicians did not
know how to handle such functions, but Riemann came up with the idea
of defining the Riemann surfaces, such that when the functions are defined
on these, they become single-valued or 'real' functions. The arithmetical
genus was defined as a topological invariant when mathematicians went
from studying two-dimensional to arbitrary-dimensional surfaces. The final
example is the introduction of sheaves. Loosely, it might be said that they
were introduced by Leray to be able to relate the cohomology of two spaces

11
Reading this description, one might think that Grothendieck necessarily needed to
introduce the ft"-group to make everything work. However, it is possible to define an-
other quotient which will make the expression of the function possible. The reason that
Grothendieck chose the group K{X) instead of this other group is presumably because
the if-group is more general.
ONTOLOGY AND MATHEMATICAL PRACTICE 253

when a function is defined between the spaces.12


A second observation I make from the case study is that, in some of
the above cases, it can be said that the new objects are constructed from
previously accepted objects by accepted construction methods. Recall that
the K-group is constructed from coherent sheaves by first taking the free
abelian group generated by these sheaves and then by forming a quotient
of this group. The arithmetical genus was constructed from holomorphic
forms defined on a complex manifold by forming the linear combination:

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


where gi is the complex dimension of the vector-space of holomorphic i-
forms on an n-dimensional manifold. Finally, a sheaf is contructed from a
collection of modules: Leray defined a sheaf B as a function that to each
closed subset F of E associates a module, BF, such that Bq> = 0 and for
/ C F, closed subsets of E, there is given a transitive restriction-map
rj : BF —> Bf. Later Cartan modified this definition, letting the modules
be associated to open subsets. When constructing a new mathematical ob-
ject, it is crucial that the construction methods give a uniquely determined
and well defined object. In the case of forming quotients, one divides the
members of the old collection into disjoint parts, and then the quotient is
the group consisting of these parts as members. Therefore, if the object
one starts out with is well defined, it is certain that the quotient also will
be well determined. Thus it can be said that mathematics consists of a cer-
tain collection of accepted objects and that new objects can be obtained by
performing accepted and well-defined constructions on the already existing
ones. This description of mathematical objects as constructed from previ-
ously existing objects will be crucial when I later describe how it is possible
to access a mathematical object and when explaining how to distinguish
between mathematical objects and fictional entities.
The above examples indicate the following two points:
1. Mathematical objects grow out of mathematical practice: mathemati-
cians formulate problems, and to be able to solve these problems they
(sometimes) need to introduce new objects.
2. New mathematical objects can be constructed from already existing
objects using accepted construction methods.
I believe that these observations indicate that mathematical objects are in-
troduced by mathematicians: it is the mathematician who formulates the
12
There is an interesting story connected to Leray's work in algebraic topology, as
described in Houzel [1998]. During the Second World War Leray was a prisoner of war
at a university in Austria. To make sure that his research would not be used by the
Germans, Leray chose to work in an area where applications were less probable than in
the area he used to work in. This made him turn to algebraic topology, and it is during
this work that he defines the notion of a sheaf.
254 CARTER

problem. In the case study the problems are based on developments in


mathematics and it is the problem that suggests the introduction of new
objects. Furthermore, the objects are obtained by making Constructions
on already existing objects. However, it is inadvisable to draw general
conclusions from a single case study. Also, if it is the case, as claimed in
realist positions, that mathematical objects exist in an independently exist-
ing realm, then, of course, it can be argued that the history of mathematics
only describes the history of our discovery of -the mathematical objects.
Therefore, to support the claim that mathematical objects are introduced

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


by mathematicians, it would seem to be necessary to argue that mathemat-
ical objects do not exist prior to their introduction. It is beyond the scope
of this paper to go into a detailed discussion of the arguments in favor of
realism.13 The objective in the rest of this paper will be to sketch a posi-
tion about the ontology of mathematics where the objects are introduced
by mathematicians. If one is a convinced Platonist, then one may read the
rest of the paper as a description of mathematical practice regardless of
whether Platonism or Anti-Platonism holds. Something like the perspec-
tive taken by Marco Panza: 'My point is simply that the mathematical
corpus, as it appears to us today, has been constructed along the history
by human beings, working in concrete and determinate situations' (Panza
[2003], p. 122).
Mark Balaguer [1998] discusses realism and anti-realism in mathematics
and finds that there are only two viable positions: Full-Blooded Platonism
(FBP) and fictionalism, but that there are no arguments that will ever let
us decide between these two. However, in the concluding chapter, Balaguer
writes:
I am in agreement with almost everything that FBP-ists and notionalists say
about mathematical theory and practice, but I do not claim with FBP-ists that
there exist mathematical objects (or that our mathematical theories are true)
and I do not claim with notionalists that there do not exist mathematical
objects (or that our mathematical theories are not true). (Balaguer [1998],
p. 179)
In what follows, we shall see how this apparent contradiction is possible.
3. Description of mathematical objects
When seeking a description of mathematical objects as introduced by ma-
thematicians, I find it useful to draw on a description in Hilbert's early
writings. Hilbert describes mathematical objects as thought-things (Gedan-
kendinge). In his famous Grundlagen der Geometrie, he writes 'We think
three different systems of things' (Hilbert [1968], p. 4, my translation). In
the lecture 'Logische Principen des mathematischen Denkens' from 1905,
Hilbert writes:
13
In Carter [2002] various realist positions are discussed.
ONTOLOGY AND MATHEMATICAL PRACTICE 255

I have the ability to think things, and to designate them by simple signs
(a,b,...,X,Y,...) in such a completely characteristic way that I can always
recognize them again without doubt. My thinking operates with these de-
signated things in certain ways, according to certain laws, and I am able to
recognize these laws through self-observation, and to describe them perfectly.
(Quoted from Peckhaus [2003], p. 150)
Thus mathematical objects are thought-things or creations of the mind,
and they exist independently of non-cognitive reality (the objective physi-
cal world). Peckhaus notes that this conception of mathematical objects

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


was not uncommon in Hilbert's time, as both Dedekind and Veronese held
a similar view. Representing the thought-things by symbols or signs is
precisely what makes the thought-things objective. Using the symbol of a
thought-thing enables mathematicians to access the thought-thing, but this
can only happen through a cognitive act of thinking the thing. Peckhaus
admits that this description is a naive approach to Hilbert's concept of
existence. One also has to take into account Hilbert's criteria of existence,
for example, existence as tied to a system and existence as following from
freeness of contradiction. Hilbert presumably was aware of his philosophical
limits and, as was typical of his style, he sent for the philosophers Nelson
and Husserl to assist him with this part of his programme.
I find Hilbert's description of mathematical objects as thought-things
very appealing. I believe that mathematical objects are created by mathe-
maticians, but that they are still objective. This, as Peckhaus explained, is
obtained through representations or symbols of the mathematical objects.
Let us consider some of the objects from the case study, to see how it is
possible to gain access to them. Our main object is Grothendieck's K-
group. The symbol of the iiT-group is K(X), but this does not tell us
anything about how to access the group. The choice of a symbol is rather
arbitrary. Instead, we need the description. The if-group is defined as a
quotient:
E{X)
K(X) =
Q{XY
where E{X) is the free abelian group of coherent sheaves defined over X
and Q(X) is a certain subgroup of E(X). To access the K-group, one needs
to know about free abelian groups, about quotients and possibly about co-
herent sheaves. If one knows about free abelian groups, then one knows
that it is possible to form such a group generated by coherent sheaves.
Also, whenever there is given a group, it is possible to form quotients of
this group, which as stated before means that one divides the elements of
the group into disjoint parts and considers, as the new group, the group
consisting of these parts. So, not only is the representation of the object
helpful, but the description of how to define or create the object is crucial
to accessing the object. It is, of course, still crucial that the object can
256 CARTER

be constructed by the given methods and that the result only yields one
object. In this case it is obvious that there is one and only one free abelian
group, and it is a standard algebraic operation to form quotients. Things
are not as easy when considering Riemann's definition of the Riemann sur-
face. Riemann merely gives an intuitive description of how the surface
looks, and because of this somewhat vague description, other mathemati-
cians struggled to understand his meaning. Clearly, it was possible to give
examples of Riemann surfaces, but the problem was to formulate a general
definition. As we saw, Weyl later succeeded in giving a precise definition

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


of Riemann surfaces and could give the subject a rigorous treatment.14 To
conclude, it is not merely the symbols that enable the mathematicians to
access mathematical objects; more important is a description of how the
object is formed, i.e., the construction of the object.15
In her 'Representational innovation and mathematical ontology' Mun-
tersbjorn proposes a position stating 'that mathematical objects are real
but emergent phenomena' ([2003], p. 160) and that mathematics is culti-
vated. Based on a study of the emergence of functions in the work of Cav-
alieri, Wallis, and Leibniz, Muntersbjorn concludes: 'This episode suggests
that novel notations are introduced (primarily) to represent operations to be
performed on extant objects. Subsequently, these novel notations may refer
to new objects as the scope of mathematical reasoning increases' ([2003],
p. 161). In my case study, one sees that the if-group is formed by per-
forming constructions on extant objects, but I believe that it is justified to
say that by performing these operations, a new object is constructed. As I
have argued above, I also hold that one should be careful to state that it is
the symbol that represents the object. A description of how the object is
formed is crucial when accessing the object. These comments aside, I am
sympathetic to Muntersbjorn's views that mathematical objects emerge
from mathematical practice, although she is not very specific about the
nature of these mathematical objects.
It is not possible to draw conclusions generally about the ontology of
mathematics from a single case study. Another and similar study has been
done by M. Leng ([2002a]). Leng has made a case study in C*-algebra,
and, although her main objective is to discuss Lakatos's Proofs and Refu-

14
That a mathematical object or concept is altered through history is common practice
in mathematics, and this is one of my reasons for believing that mathematical objects are
introduced by human beings. At any given time, we acknowledge certain objects as the
ones we are interested in studying. But we are still able to consider former definitions or
descriptions of these objects. Some realists would state that our changing definitions of
mathematical objects provide us with better and better descriptions of the mathematical
objects that we are trying to describe.
15
Sometimes when the object is described by standard notation the description is not
needed. However, it is common practice for books and articles in mathematics to start
out by presenting the notation that will be used, even though much of it may be standard.
ONTOLOGY AND MATHEMATICAL PRACTICE 257

tations, she also proposes an answer to the question about the status of
mathematical objects. Interestingly, her conclusion is that mathematical
objects are not essential for mathematicians as they are more interested in
the arguments that they are able to give for the claims that they make.
Thus, the position that Leng proposes is an anti-realist position with re-
spect to the existence of mathematical objects. Leng, however, is not all
that clear on what she means by anti-realism. If anti-realism is supposed
to mean that mathematical objects do not exist independently of human
beings, then obviously I agree with her. But I do hold that it makes sense

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


to speak about objects in mathematics and, as we are able to speak success-
fully about them, they are entitled to some kind of existence. My objective
is to find a satisfactory account of the existence of mathematical objects.
Concerning the status of mathematical objects, I claim that they are
objectively accessible, which follows from the rigorous descriptions that
mathematicians are able to give them. Here I distinguish between exis-
tence and accessibility. A mathematical object exists when it has been
constructed by a human being, but it has to be presented somehow to one
or several members of the mathematical community to be accessible. But
what about the nature of the mathematical objects? Are they merely mul-
tiple ideas in the mathematicians' minds? To give a well argued answer
to this question is not easy. However, I do believe that it is possible to
argue that mathematical objects exist as a kind of an abstract object after
they have been introduced or, put differently, that they may be regarded
as inhabitants of Popper's third world (see Popper [1972]). I will hasten
to add that, when I use the word 'abstract', I do not mean abstract in the
usual sense, where an object is abstract if it exists outside time and space.
A more appropriate distinction between abstract and concrete, I believe, is
given by: Abstract entities are human constructions or creations, whereas
concrete entities exist independently of human mind or language.
An argument for the general existence of abstract objects can be found by
using the fact that human beings can successfully communicate. One may
ask what it is that enables us to do that. The most obvious explanation16
is that the concepts that we use are of objectively existing abstract objects.
A further observation supporting this argument is the fact that mathe-
maticians usually work together when doing mathematics. It would seem
strange to think that they each are creating their own objects, when they
can agree completely on all the details. Drawing again on Hilbert's writings,
in this case his description of our knowledge of the finite part of mathema-
tics (Hilbert [1996]), might give an idea of how objects can be said to exist
objectively. Hilbert writes that
16
Of course we could say that communication is not possible, that it is just something
we imagine, but then we would be skeptics and all, realists and anti-realists alike, be in
the same boat—not knowing anything.
258 CARTER

the objects of number theory are for me—in direct contrast to Dedekind and
Prege—the signs themselves whose shape can be generally and certainly recog-
nized by us—independently of space and time, of the special conditions of the
production of the sign, and of insignificant differences in the finished product.
(Hilbert [1996], p. 1121)
The most sensible way of understanding this statement is to take Hilbert to
mean that the mathematical object is the type of the physical inscriptions,
i.e., when he writes that the mathematical objects are the signs themselves,
he means that the mathematical object is the type of the signs or symbols

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


that we use to represent them with. Then it makes sense to say that we are
always able to recognize them independently of, for example, the production
of the sign. Similarly, we can regard the single mathematician's creations
or the symbols that are used to represent an object, as tokens and thus
the objective mathematical object is the type of all these tokens. As we
clearly have the ability to distinguish between types and tokens, I think
that this interpretation gives a good picture of what goes on.17 Below, we
shall see an alternative way of arguing that a mathematical object exists
as a genuine object.

4. Mathematical Objects and Fictional Entities


If mathematical objects are introduced by mathematicians, it may be ar-
gued that they are mere fictions. Thus, my next aim is to explain the
difference between mathematical objects and fictional entities such as the
characters appearing in a play, for example, Hamlet. This will be dealt with
by addressing two questions: 1) What distinguishes fictional entities and
mathematical objects? 2) If mathematical objects are created by mathe-
maticians, then how do we explain that mathematics is useful to the natural
sciences?
The latter question is a reformulation of the indispensability argument,
where it is argued, from the indispensability of mathematical objects to our
best theories about the world, that they exist. There are several ways of
dealing with this question. One approach is to reject this argument which is
regarded by some philosophers as the most successful argument for realism
in mathematics. Indeed, H. Field [1980] wrote that the indispensability ar-
gument is the only serious argument for realism in mathematics. Although
there are still some proponents of an indispensability type of argument (for
example, Resnik [1997], and Colyvan [2001]) I believe that recent criticism
17
Note that this picture extends the type/token distinction. I have no explanation of
how this distinction may be extended to conceptual objects; it just seems reasonable
to assume that it is the same faculty that we use. When mathematicians discuss a
mathematical object, they seem to be talking about a single object, although they may
even use different representations for it. A simple example is TT. The number n can be
thought of as the ratio between the circumference and the diameter of a circle or as a
number appearing in the expression of a sum of a certain infinite sequence.
ONTOLOGY AND MATHEMATICAL PRACTICE 259

of the argument (see for example P. Maddy, [1992], S. Feferman [1998] and
[2000], and finally M. Leng [2002b]) speaks in favour of its rejection. My
own reason for rejecting the argument is because it is not true to the prac-
tice of mathematics. Elsewhere (Carter [2002]), I argue that the argument
is untenable mainly because, when the existence of mathematical objects is
judged from whether they are used in a scientific theory or not, then some
objects exist while others do not, even though the non-existent objects can
be constructed from the existing ones. Colyvan [2001] argues that the ob-
jects of a mathematical theory could be taken to exist as long as there is

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


a series of mathematical theories, where the mathematical theory they are
defined in is used in the next mathematical theory, and so on, until the last
mathematical theory is used in a scientific theory:
On a holistic view of science, even the most abstract reach of mathematics
is applicable to empirical science so long as it has applications in some other
branch of mathematics, which may in turn have applications in some further
branch until eventually one of these finds applications in empirical science.
Indeed, once put this way, it is hard to imagine what part of mathematics
could possibly be unapplied. (Colyvan [2001], p. 107)
However, considering mathematical practice, there seem to be parts of
mathematics which are defined without the intent of being applied—and
certainly not in a scientific theory. (Compare also footnote 12 about Leray
deliberately choosing to work in a branch of mathematics where he hoped
no applications could be found.) The main result in the case study is
the theorem of Grothendieck-Riemann-Roch. This theorem is a generali-
zation of Hirzebruch-Riemann-Roch, which was the culmination of all the
attempts to find a generalization for arbitrary dimensions of the original
theorem of Riemann and Roch. Grothendieck's motivation to find a new
generalization was his focus on abstract algebraic notions, i.e., he wanted
to find a generalization valid for all fields (instead of the complex num-
bers), and he wanted to find a version where the focus was on morphisms
instead of objects, which reflects the emergence of categorical notions in
mathematics. The matin purpose of proving the theorem of Grothendieck-
Riemann-Roch was not that it had potential applications either in science
or in mathematics. The main purpose was to formulate and prove a version
of the theorem using all the new available machinery.
When discussing the indispensability argument it is essential to consider
what happens when mathematics is used in science. In Platonism and Anti-
Platonism in Mathematics Balaguer also adresses this question, formulating
the principle (TA): 'Empirical theories use mathematical-object talk only
in order to construct theoretical apparatuses (or descriptive frameworks)
in which to make assertions about the physical world' (Balaguer [1998],
p. 137). Balaguer gives the following example: the statement 'The physical
system 5 is forty degrees Celsius' does not have to refer to an independently
260 CARTER

existing number forty to be true. Balaguer puts this as


Isn't it just obvious that the only reason we refer ... to the number 40 is that
this provides us with a convenient way of saying what S's temperature state
is? What the Celsius scale does is correlate different temperature states with
different numbers, so that numerals can serve as names of temperature states,
(p. 138)
We have seen in the case study that mathematical entities are introduced
into mathematical theories, and it would then be the mathematical theories
that are applied to the scientific theories. It therefore would seem that when

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


mathematics is used in scientific theories, it is the theorems that are used,
not the entities in the theorems. When you apply Pythagoras's theorem
to a triangle, you say that the square of the hypoteneuse of your triangle
is equal to the sum of the squares of the other two sides. You say nothing
about an abstract triangle and whether it exists. It is when you deduce
theorems about the triangle that you refer to the abstract triangle. When
chemists use a group of symmetries to describe the structure of a molecule,
it is the symmetries of the molecule that have the structure of a group;
they do not postulate new objects that are members of the group. So we
could say that group theory is about this molecule. Of course there are
many molecules, and they all share this structure. This can be expressed
as: the molecules are tokens of this structure, whereas the structure that
they share, the group, is the type. But the question now is: Does the
indispensability argument give existence to the type?
Rejecting the argument is not the solution, however, to explaining why
mathematics is useful. Part of the explanation is that mathematics is intro-
duced to be useful (originally in dealing with the physical world, but later
also to mathematics itself). The history of mathematics shows that for a
long time its development was intertwined with science. Even today, sci-
ence is a source of inspiration for mathematics.18 However, it is remarkable
how mathematical theories which have been developed without the intent of
being used in the natural sciences sometimes do find application. This has
been expressed in Wigner's 'The unreasonable effectiveness of mathematics
in natural science' [I960]. I must admit that I have no explanation for this
apparent mystery, but with regard to it, I believe that both realists and
anti-realists are in the same boat: if mathematical objects exist as Platonic
entities outside space and time, then it is just as much a mystery that these
obects have anything to say about our theories about the physical world,
as it is a mystery that mathematics, which is introduced by human beings,
is useful. After all, there must be a reason for introducing mathematics.

18
A while ago, I heard Michael Atiyah talk about a problem which had been proposed
to him by a physicist. Actually the problem turned out to be very hard to solve, and as
Atiyah had not yet found a solution, he promised a bottle of champagne to anyone in
the audience who could provide one.
ONTOLOGY AND MATHEMATICAL PRACTICE 261

We now turn to the first question about what distinguishes a fictional


entity from a mathematical object. Here I will draw on a distinction in-
troduced by Marco Panza (Panza [2003]). By exploiting the concepts of
a definition and the act of exhibition, Panza is able to draw a distinction
between genuine objects and fictions:
The distinction between definition and exhibition is crucial. It does not hold
for all sorts of objects. Rather we could distinguish in general between two
sorts of objects: The objects which can be both be defined and exhibited
(these two acts being different one from the other) and the objects that can

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


only be defined—or for which there is no possible distinction between the act
of definition and the act of exhibition. (Panza [2003], p. 122)
Thus one could ask the question: 'Which is the highest mountain in the
word?' and the answer to this question would be Mount Everest. This
object can be exhibited by an ostensive act. One could also ask the question:
'Who is the doctor who helps Sherlock Holmes?' and the answer would be
'Dr. Watson'. However, in this case there is no way of exhibiting the
character 'Dr. Watson' other than giving further descriptions from the
books. With regard to mathematical objects, Panza's point is that the act
of exhibiting an object does not necessarily have to consist in exhibiting an
empirical object which the definition19 is supposed to apply to. Thus he
writes:
According to my definition of mathematics, mathematical objects are genuine
objects. This means that you are doing mathematics only if the objects you
are dealing with are such that you have a critierion to distinguish the act of
exhibiting them, from an act of defining them. To introduce mathematical
objects then you have to define them in some way, and to fix the conditions
under which such objects are exhibited. (Panza [2003], p. 125)
Panza suggests that empirical objects in some cases may represent mathe-
matical objects in a suitable way; for example, a triangle may be represented
by a drawn triangle. This is not always possible, however. In the case study
presented in this paper it is not obvious what such representations would
look like. Thus we have to look for other ways of describing the act of
exhibiting these mathematical objects. (Note that when a mathematical
theory finds an application, then there will be a way of representing the
mathematical objects in the theory.) Let us return to the case study, to
consider what might count as an act of exhibition of some of the objects
that we have encountered there. We shall see that the acts of exhibition are
derived from the methods of construction which we have already seen also
provide accessibility to the objects. Recall the K-group, which was defined

19
Panza distinguishes between different sorts of definitions: A definition may charac-
terise an Object, a class of objects, or a certain sort of object.
262 CARTER

by the expression:
E{X)
K{x)
~
where E(X) is the free abelian group generated by isomorphism classes of
coherent sheaves over X and Q(X) is the subgroup generated by expressions
of the type
T-T' -T"
whenever there is a short exact sequence

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


of coherent sheaves over X.
This is the definition of the if-group. The question is what would count
as an act of exhibition of the group. Above, I stated that new objects
are constructed from already accepted ones. On this line of thought, it is
reasonable to suppose that the objects that the if-group is formed by, i.e.,
coherent sheaves, are genuine objects. Then to ensure that the if-group is
a genuine object consists in proving the following:
• A free group generated by genuine objects is also a genuine object;
• A quotient of a genuine object is also a genuine object.
To prove the above, one has to prove that it is possible to form a free abelian
group from a set of objects and secondly, that when given an equivalence
relation, it is possible to form a quotient. To form a free abelian group and
a quotient are standard mathematical operations; thus the only thing that
remains to be done in this case is to prove that Q{X) defines an equivalence
relation.
Panza also introduces another distinction which can be used to charac-
terize mathematical objects:
I call 'authentic' the genuine objects which are such that their exhibition pro-
vides a content that can be considered as a source to ascribe some properties
about these that are not ascribed to them by a tautology about them. (Panza
[2003], p. 127)
For example, the if-group has the trivial property following directly from
the definition of being a group, that it has an associative binary operation
defined on it. An example of a property which is non-trivial is that it is
possible to prove that the if-group can be given the structure of a ring.
This proof uses that there can be defined an isomorphism between the K-
group and another quotient, KX{X), which can be obtained from a free
abelian group generated by vector bundles. Using this isomorphism, it is
possible to transfer the ring structure defined on Ki(X) to K{X). Part
of the definition of a quotient G/N is a map h : G -> G/N that maps
a member of G to its equivalence class in the quotient group. This map
ONTOLOGY AND MATHEMATICAL PRACTICE 263

enables the mathematician to work with the representations in G instead of


the equivalence classes. In the above proof these maps are also used to work
with (linear combinations of) sheaves and (linear combinations of) vector
bundles. Considering this map as part of the exhibition of the -ff-group,
one sees that this exhibition is part of the reason it is possible to deduce the
theorem. Furthermore, assuming that a sheaf is a genuine object, then this
map provides a method of exhibiting the members of the quotient group.
A final example is the expression of the morphism

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


i=0

This is a definition of the morphism. To prove that this morphism exists,


or equally to give the criteria of the acts of exhibition of this entity, consists
in proving that: i) the objects Rlf*{!F) exist, and ii) there are only a finite
number of these that are different from zero (making certain that the sum
is finite). That the sheaves Rlf*(J-) exist follows from the fact that the
sheaf T is an abelian category. That there only exist a finite number of
these that are different from zero follows because the sheaf is coherent and
the map / is proper, as remarked in section 2.2.
To sum up, the picture of mathematical objects arising from the case
study is that they are introduced and constructed by mathematicians. Be-
cause of the unique description the mathematician is required to make when
introducing an object, it is objectively accessible. Furthermore, it is also
possible to distinguish the acts, as described by Panza, of defining an object
from the act of exhibiting an object, making them genuine objects. Thus
the practice of mathematics at any given time can be regarded as based on
a set of accepted objects and a set of accepted construction methods. New
objects can be introduced by applying the accepted construction methods
to the existing mathematical objects.20
Above I have presented the position that I believe comes out of the case
study in K-theory. Although there is no general agreement among those
who study the practice of mathematics, it seems to me that we at least agree
in rejecting the traditional platonist account of mathematics in favour of a
view of mathematics as somehow created.

4. Appendix
In this appendix some basic notions of algebraic geometry will be defined.
These include algebraic varieties and sheaves.
20
Note that not all objects are introduced in this way. Elsewhere (Carter [2002]) I
discuss another category of objects which I denote the disputed objects. Examples of
these are the infinitesimals, the choice function, and inaccessible cardinals.
264 CARTER

Algebraic varieties can be thought of as the common set of zeros of a


given family of polynomials. An attractive feature of algebraic varieties is
that it is possible to study geometrical objects by studying the algebraic
properties of the polynomials that define them.
A more precise definition is: Let A; be an algebraically closed field of
characteristic 0 and let A^ = {ai,...,a n } for a* € A; be the affine space of
n-tuples of elements of k. k[xi,.. .,xn] denotes the ring of polynomials in
the variables xi,...,xn with coefficients in k. Given a family of polynomials
/* G k[xi,...,xn], i = 1,2,..., an algebraic variety is defined as the set

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


V = {P&An: p{P) = 0 for all i}.
Algebraic varieties are usually studied in projective space. The projective
n-space, Pn, over k is defined as the quotient of the set of (n + l)-tuples
{ao,..., a n }, the ajS not all zero, by the equivalence-relation {a0, • • •, an} ~
{bo, - •., bn} if and only if there exists a A € k\{0} such that bi = AOJ for
alH = 0,.. .,n.
Now, because of this equivalence-relation, the definition of an algebraic
variety on projective space needs some modification.
If a point {a0,... ,an} is a zero of the function / then any point
{Aao> • • • j ^&n} should also be a zero of / . Choosing / to be a homogeneous
polynomial, i.e., a polynomial where the degree of all its monomials is con-
stant, say m, this will be the case, as /(Aa 0 ,. •., Xan) = A m /(a 0 ,..., a n ).
A projective algebraic variety is then defined as the set W = {T €
pn . y ( j ' ) = 0 for all i}, for /*, a family of homogeneous polynomials in
k[xo,...,xn].
A sheaf of abelian groups over a topological space X associates to each
point p of X an abelian group, where the group-operations are subject to
some 'continuity' conditions. Defining sheaves over X, one may start by
defining pre-sheaves over X:
To each open set U is associated an abelian group, F{U), and for U CV,
open sets in X, there is a restriction map p\j : T(V) -> F{U) subject to
the following conditions:

0)
(ii) P^ j { u )
(iii) For U CV CW, open sets in X, we have that p™ = Py ° P'
Now for each point p in X one takes the direct limit of the groups
for all open sets Ui containing p, with respect to the restriction maps py.
This gives groups Tp and canonical maps p% : F{U) -> Tv for each U
containing p. The collection of Tp, as p runs through X, is called a sheaf
of abelian groups over X. Thus J-' — Up!Fp.
A sheaf is usually thought of as a triple (X, TT,T), where TT : T -> X is a
continuous map. It is possible to define a topology on T, which is induced
from the topology on X. Then one can define the map n as taking an
ONTOLOGY AND MATHEMATICAL PRACTICE 265

element / e Tv to the element p i n X. Sheaves can also be defined in a


similar fashion as sheaves of rings over X and as sheaves of ^-modules over
X, whenever A is a sheaf of rings over X.
Grothendieck worked with a special kind of sheaves, namely coherent
sheaves. This notion was first defined by Cartan in 1950 and the theory
about them was further developed in Serre's paper (Serre [1955]).
Given a sheaf (X,Tr,!F) over X and an open subset U of X, one defines
sections T(U, J) over U as the set of maps s : X ->• ^"such that TTOS :U -> U
is the identity-map on U. This set also forms an abelian group, and if U

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


runs through all open sets of X, we obtain a presheaf over X.
Definition 1. A sheaf of >l-modules, T, is of finite type if it is locally
generated by a finite number of its sections.
Definition 2. A sheaf of A-modules, T, is coherent if
(a) T is of finite type;
(b) si,...,sp are the sections of Tover an open set U then the kernel
of the map <p : A*(U) ->• T(U) defined by taking {fu ..., fp} £ AP to the
element X))=i fisi(x) m ?x is of finite type.
A consequence of this definition is that locally a coherent sheaf is isomorphic
to a sheaf-quotient of Aq for some q. One may think of a coherent sheaf as
a finitely generated sheaf.
266 CARTER

References
ATIYAH, M., and F. HIRZEBRUCH [1988]: 'Vector bundles and homogeneous
spaces', in M. Atiyah, Collected Works. Vol. 2. Oxford: Clarendon Press,
p. 53-84.
BALAGUER, M. [1998]: Platonism and Anti-Platonism in Mathematics. New
York: Oxford University Press.
BOREL, A., and J.-P. SERRE [1958]: 'Le theoreme de Riemann-Roch', Bull. Soc.
Math. France 86, 97-136.
CARTER, J. [2002]: Ontology and Mathematical Practice. Ph.D. thesis, Odense:

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


University of Southern Denmark.
CARTIER, P. [2000]: 'Grothendieck et les motifs', in Notes sur l'Histoire et la
Philosophie des Mathematiques IV, IHES/M/00/75, Novembre, pp. 1-36.
COLYVAN, M. [2001]: The Indispensability of Mathematics. New York: Oxford
University Press.
CORFIELD, D. [2003]: Towards a Philosophy of Real Mathematics. Cambridge:
Cambridge University Press.
EILENBERG, S., and MAC LANE, S. [1942]: 'Natural isomorphisms in group the-
ory', Proc. Natl. Acad. Sci. USA 28, 537-543.
[1945]: 'General theory of natural equivalences', Trans. Amer. Math.
Soc. 58, 231-294.
FEFERMAN, S. [1998]: 'Why a little bit goes a long way: Logical foundations
of scientifically applicable mathematics', in In the Light of Logic. New York:
Oxford University Press, pp. 284-298.
[2000]: 'Relationships between constructive, predicative and classical
systems of analysis', in V. F. Hendricks, S. Andur Pedersen, and K. F. J0rgen-
sen, eds., Proof Theory: History and Philosophical Significance. Dordrecht:
Kluwer Academic Publishers, pp. 221-236.
FIELD, H. [1980]: Science without Numbers. Oxford: Basil Blackwell.
GROSHOLZ, E., and H. BREGER, eds. [2000]: The Growth of Mathematical
Knowledge. Dordrecht: Kluwer Academic Publishers.
GROTHENDIECK, A. [1956/57]: 'Sur les faisceaux alge'briques et les faisceaux
analytiques coherents', Se~minaire H. Cartan. Secretariat math&natique, Ecole
Normale Superieure, Paris, No. 2.
[1971]: 'Classes de faisceux et theoreme de Riemann-Roch', in Seminaire
de GSomitrie Alge'briques du Bois Marie 1966/67. Berlin and New York:
Springer Verlag, pp. 20-77.
th
HILBERT, D. [1968]: Grundlagen der Geometrie (10 ed.). Stuttgart: B. G.
Teubner.
[1996]: 'The new grounding of mathematics', in W. B. Ewald, ed., From
Kant to Hilbert. A Sourcebook in the Foundations of Mathematics. New York:
Oxford University Press, 1115-1134.
HIRZEBRUCH, F. [1966]: Topological Methods in Algebraic Geometry. New York:
Springer Verlag.
HOUZEL, C. [1998]: 'Histoire de la theorie des faisceaux', Simin. Congr. 3. Paris:
Soc. Math. France, 101-119.
LAKATOS, I. [1976]: Proofs and Refutations: The Logic of Mathematical Discov-
ery. Cambridge: Cambridge University Press.
ONTOLOGY AND MATHEMATICAL PRACTICE 267

LARVOR, B. [2001]: 'What is dialectical philosophy of mathematics', Philos.


Math. (3) 9, 212-229.
LENG, M. [2002a]: 'Phenomenology and mathematical practice', Philos. Math.
(3) 10, 3-25.
[2002b]: 'What's wrong with indispensability? (Or, the case for recre-
ational mathematics)', Synthese 131, 395-417.
LERAY, J. [1946]: 'L'anneau d'homologie d'une representation, C. R. Acad. Sci.
Paris 222, 1366-1368.
MADDY, P. [1992]: 'Indispensability and practice', J. Phil. 89, 275-289.

Downloaded from http://philmat.oxfordjournals.org/ at University of Bath on June 18, 2015


MARQUIS, J.-P. [1997]: 'Abstract mathematical tools and machines for mathe-
matics', Philos. Math. (3) 5, 250-272.
MUNTERSBJORN, M. [2003]: Representational innovation and mathematical on-
tology. Synthese 134, 159-180.
PANZA, M. [2003]: 'Mathematical proofs', Synthese 134, 119-158.
PECKHAUS, V. [2003]: 'The pragmatism of Hilbert's programme'. Synthese 137,
141-156.
POPPER, K. [1972]: Objective Knowledge. An Evolutionary Approach. Oxford:
Oxford University Press.
RESNIK, M. [1997]: Mathematics as a Science of Patterns. New York: Oxford
University Press.
RIEMANN, B. [1892]: 'Theorie der Abelschen Functionen', in H. Weber, ed.,
Bernhard Riemann Gesammelte mathematische Werke und wissenschattlicher
Nachlass. Leipzig: B. G. Teubner.
ROCH, G. [1865]: 'Ueber die Anzahl der willkiirlichen Constanten in algebraischen
Functionen', J. fur Math. 64, 372-376.
SERRE, J.-P. [1955]: 'Faisceaux algebriques coherents', Ann. of Math. 61, 197-
278.
WEYL, H. [1913]: Die Idee der Riemannschen Flk'che. Leipzig and Berlin: B. G.
Teubner.
WIGNER, E. P. [I960]: 'The unreasonable effectiveness of mathematics in the
natural sciences', Communications in Pure and Applied Mathematics 13, 1-
14.

ABSTRACT. In this paper I propose a position in the ontology of mathematics


which is inspired mainly by a case study in the mathematical discipline if-theory.
The main theses of this position are that mathematical objects are introduced by
mathematicians and that after mathematical objects have been introduced, they
exist as objectively accessible abstract objects.

You might also like