Alternatives to the Calculus:
Nonstandard Analysis and Smooth Infinitesimal Analysis
A thesis presented to
the faculty of
the College of Arts and Sciences of Ohio University
In partial fulfillment
of the requirement for the degree
Master of Arts
Jesse P. Houchens
May 2013
© 2013 Jesse P. Houchens. All Rights Reserved. 2
This thesis titled
Alternatives to the Calculus:
Nonstandard Analysis and Smooth Infinitesimal Analysis
by
JESSE P. HOUCHENS
has been approved for
the Department of Philosophy
and the College of Arts and Sciences by
Philip W. Ehrlich
Professor of Philosophy
Robert Frank
Dean, College of Arts and Sciences 3 ABSTRACT
HOUCHENS, JESSE P., M.A., May 2013, Philosophy
Alternatives to the Calculus: Nonstandard Analysis and Smooth Infinitesimal Analysis
Director of Thesis: Philip W. Ehrlich
We attempt to clarify and evaluate what shall be called Mac Lane’s thesis—the
thesis that nonstandard analysis (NSA) and smooth infinitesimal analysis (SIA) are
alternatives to the standard approach to the calculus. In doing so, we outline the
historical approaches to the calculus, the standard approach to the calculus, and two
nonstandard approaches, namely NSA and SIA; we also attempt to clarify and evaluate a
set of comparisons of NSA and SIA, namely Bell’s 5 mathematico-philosophical
contentions and Bell’s historical contention. 4
For my parents who continually remind me
“While it may not always be easy,
it will always be worth it.” 5 ACKNOWLEDGMENTS
First, I would like to thank the members of my thesis committee. Todd Eisworth provided insightful comments from the perspective of a working mathematician. In addition, his set theory course allowed me to clarify many of the ideas underlying this thesis. Stewart Shapiro travelled down to Athens for the defense as well as allowed me to participate in his in his seminar on continuity and infinitesimals at The Ohio State
University. His comments were very helpful in clarifying many of my arguments. In addition, it was in his seminar that I was first exposed to NSA and SIA and that many of the questions raised in section II.A were first voiced.
Second, I would like to express my sincere gratitude to Philip Ehrlich for advising this project. Each week, as I returned from Dr. Shapiro’s seminar, he would ask me what was discussed and my thoughts on the matter. When coming up with the topic of this thesis, he gave me invaluble advice for making the project manageable. In addition, he directed me to excellent sources on the history of the calculus as well as on each of the mathematical theories. Finally, his patience as I attempted to understand each of the theories as well as his willingness to always correct my mistakes and provide a surplus of comments on previous drafts was absolutely beneficial. Overall, one could say that
Phil’s guidance contributed significantly in improving the quality of this paper.
Finally, I would like to thank my family, former professors, and fellow graduate students as they listened to my struggles and constantly voiced their confidence in my abilities. 6 TABLE OF CONTENTS
Abstract...... 3
Dedication...... 4
Acknowledgements...... 5
List of Figures...... 7
Introduction...... 8
1. Historical and Theoretical Background...... 11 1.1. Historical Approaches to the Calculus...... 11 1.2. The Standard Approach to the Calculus...... 21 1.3. Nonstandard Analysis...... 30 1.4. Smooth Infinitesimal Analysis...... 47
2. Clarifications and Evaluations...... 57 2.1. Bell’s 5 Mathematico-Philosophical Contentions...... 57 2.2. Bell’s Historical Contention...... 70 2.3. Mac Lane’s Thesis...... 73
Conclusion...... 79
Works Cited...... 80 7 LIST OF FIGURES
Figure 1. Newton’s differential calculus………………….……………….……….13
Figure 2. Newton’s integral calculus...…………………………………………...... 14
Figure 3. Leibniz’ calculus………..………………………………………………..15
Figure 4. Toricelli’s trumpet………………………………………………………..21
Figure 5. Infinitesimal difference between dy /x and y /x …...…………...…..44
Figure 6. Motivation for nondegeneracy………….……………………….………..58
Figure 7. Local straightness SIA………………..…………………………………..62
Figure 8. Local straightness NSA 1………………………..………………………..63
Figure 9. Local straighntess NSA 2……………………………..…………………..64
8 INTRODUCTION
After a historically-sensitive discussion of integration and differentiation in
Mathematics, Form and Function, Saunders Mac Lane concludes with the following:
This discussion of alternative views of the calculus may serve to support our thesis that Mathematics is not just formalism and is not just empirically convenient ideas, but consists in formalizable intuitive or empirical ideas. Calculus fits this thesis as to the nature of Mathematics because it starts from problems (such as the calculation of areas or of rates of change) and from these problems develops suggestive ideas which ultimately can be fully formalized… Another thesis asserts that the same intuitive idea can be variously formalized. For the calculus, we know that this is the case. Indeed, the earlier and originally wholly speculative use of infinitesimals can be rigorously formalized in at least two different ways: By using Abraham Robinson’s non-standard model of the reals, as presented for example in Keisler’s text [1976] on calculus, or by using Lawvere’s proposals to use ‘elementary topoi’ in which the real line R is presented not as a field but as a ring in which there is a suitable infinitesimal neighborhood of 0 (see Kock [1981]). (Mac Lane 155)
The passage above contains two theses: A) The historical development of the calculus consisted of different attempts to formalize intuitive ideas suggested by basic problems;
B) There are at least two different ways to rigorously formalize the “originally wholly speculative use of infinitesimals”, specifically the ways employed in nonstandard analysis (NSA) and smooth infinitesimal analysis (SIA)1. These two theses, we believe,
motivate a third, namely that NSA and SIA are two alternatives to the calculus. Let this
third thesis be called Mac Lane’s thesis. John L. Bell has made a remark similar to Mac
Lane’s thesis restricted to SIA:
“As we show in this book, within [SIA] the basic calculus...can be developed along traditional ‘infinitesimal’ lines—with full rigor—using straightforward calculations with infinitesimals in place of the limit concept.” ([2008] 4)
1 While Mac Lane references Kock’s book on synthetic differential geometry (SDG), the fragment of SDG that addresses the calculus is SIA. 9 The purpose of this essay is to explicate and evaluate Mac Lane’s thesis. This will involve first discussing some of the differing historical approaches to the calculus as well as the problems that motivated its development. After doing so, the standard ε—δ approach to the calculus as well as the two nonstandard approaches, NSA and SIA, will be outlined to facilitate comparison among the theories.
According to John L. Bell, some of the differences between NSA and SIA are the following:
1. In models of [SIA], only smooth maps between objects are present. In models of nonstandard analysis, all set-theoretically definable maps (including discontinuous ones) appear. 2. The logic of [SIA] is intuitionistic, making possible the nondegeneracy of the microneighbourhoods ∆ and Mi, i = 1, 2, 3. The logic of nonstandard analysis is classical, causing all these microneighborhoods to collapse to zero. 3. In [SIA], the Principle of Microaffineness entails that all curves are ‘locally straight’. Nothing resembling this is possible in [NSA]. 4. The property of nilpotency of the microquantities of [SIA] enables the differential calculus to be reduced to simple algebra. In [NSA] the use of infinitesimals is a disguised form of the classical limit method. 5. In any model of [NSA] R* has exactly the same set theoretically expressible properties as R does: in the sense of that model, therefore, R* is in particular an Archimedean ordered field. This means that the ‘infinitesimals’ and ‘infinite numbers’ of nonstandard analysis are so not in the sense of the model in which they ‘live’, but only relative to the ‘standard’ model with which the construction began. That is, speaking figuratively, a ‘denizen’ of a model of nonstandard analysis would be unable to detect the presence of infinitesimals or infinite elements in R*. This contrasts with [SIA] in two ways. First, in models of [SIA] containing invertible infinitesimals, the smooth line is non-Archimedean in the sense of that model. In other words, the presence of infinite elements and (invertible) infinitesimals would be perfectly detectable by a ‘denizen’ of that model. And secondly, the characteristic property of nilpotency possessed by the microquanities of any model of [SIA] (even those in which invertible infinitesimals are not present) is an intrinsic property, perfectly identifiable within that model.2 ([2008], 112)
2 In Bell [2006] some of the above contentions are slightly altered. I will take these altered versions into account in my evaluation. 10 Call the above Bell’s 5 mathematico-philosophical contentions. While these contentions are helpful for comparing NSA and SIA, some are in need of clarification. For example, in contention (3) Bell describes all curves in SIA as ‘locally straight’ and claims “nothing resembling this is possible in NSA.” Yet, no definition is given for ‘locally straight’, moreover nowhere else is this term used in [1998; 2008]. This unclarity makes it difficult to understand just what it is that is impossible in NSA.
Furthermore, Bell concludes the above contentions with the following:
The differences between [NSA] and [SIA] may be said to arise because the former is essentially a theory of infinitesimal numbers designed to provide a succinct formulation of the limit concept, while the latter is, by contrast, a theory of infinitesimal geometric objects, designed to provide an intrinsic formulation of the concept of differentiability. (Bell [2008] 112)
Call this Bell’s historical contention. Laugwitz, in the following passage, challenges
Bell’s historical contention:
It is a misleading simplification to say, as the author does, that NSA “is essentially a theory of infinitesimal numbers designed to provide a succinct formulation of the limit concept…” (Laugwitz [1999] 2).
Motivated by this challenge as well as the previously stated unclarities, we will attempt to clarify and evaluate Bell’s 5 mathematico-philosophical contentions and Bell’s historical contention before examining Mac Lane’s thesis, determining whether it is collectively illustrated by NSA and SIA.
11 1. HISTORICAL AND THEORETICAL BACKGROUND
This section is intended to provide the relevant historical and theoretical background for the clarification and evaluation of Bell’s 5 mathematico-philosophical contentions, Bell’s historical contention, and Mac Lane’s thesis. Section I.A presents some of the basic problems that motivated the development of the calculus as well as the historical approaches to solving such problems. Section I.B outlines the standard ε—δ approach to the calculus following Binmore [1982]. Section I.C outlines NSA following the approach presented in Keisler [2007]. And, section I.D outlines SIA following (for the most part) the approach presented in Bell [1998; 2008].
1.1. Historical Approaches
Some of the problems that motivated the development of the calculus include: (1) finding the instantaneous velocity of an object in continuous motion; (2) finding the area under a continuous curve; and (3) rigorously describing the continuity of functions.
While the ancients and medievals present historically interesting approaches to solving cases of (1) and (2) (c.f. Baron [1969]), we focus our discussion on the historical approaches beginning in the 17th century with Newton and Leibniz. We will first
overview the calculi of Newton and Leibniz. After which, we will discuss some of the
attempts in the 18th and 19th centuries to make these approaches more rigrorous.3
Before outlining the historical approaches of Newton and Leibniz, we note two
distinctions Guicciardini has made between their calculi and the contemporary calculus:
3 We note that if the reader is interested in more comprehensive accounts of the history of the calculus, it is recommended that they see, among other works, Boyer [1949], Edwards [1979], Guicciardini [2003], and Jahnke [2003]. 12 (a) “Neither Newton’s nor Leibniz’s calculi are about functions;” instead, they “talk in terms of ‘quantities’... and they refer to these quantities, their rates of change, their differences, etc. related to specific geometric entities” (Guicciardini [2003] 73). (b) “While we are used to referring to calculus as the continuum of real numbers, the continuum to which Newton and Leibniz refer is geometrical or kinematical” (Guicciardini [2003] 73-74).
With these distinctions in mind, let us first look at the calculus of Newton.
Methods for finding the tangent of a curve or the instantaneous velocity of an object in continuous motion were developed and employed by a number of predecessors of Newton and Leibniz including Fermat and Barrow.4 Newton’s approach employs
what he referred to as ‘fluents’, ‘fluxions’, and ‘moments’; respectively, “the quantities
v,x,y,z generated by a ‘flow’,” the instantenous speeds vÝ ,x Ý , yÝ ,z Ý , and “the infinitely small additions,o, by which those quantites increase during each infinitely small
interval of time” (Guicciardini [2003] 78). In the passage below, Guicciardini explains
the ways that these quantities were employed for differentiation. 5
Imagine a plane curve f (x,y) 0 to be generated by the continuous flow of a point P(t). If x,y are the Cartesian coordinates of the curve, yÝ / xÝ will be equal to tan , where is the angle formed by the tangent in P(t) with the x-axis (see Fig. 3.3). According to Newton’s conception, the point will move during the “indefinitely small period of time” with uniform rectilinear motion from P(t) to Pt o . The infinitesimal triangle indicated in Fig. 3.3 has sides equal to yÝ o and xÝo and so tan yÝo /xÝo yÝ / xÝ . An extremal point will have yÝ / xÝ tan 0. (Guicciardini [2003] 79)
4 C.f. Edwards pp. 122-141 5 We note that in this passage, Guicciardini uses functions as a convention to represent variable quantities. As stated in (a), neither Leibniz nor Newton speak of functions. 13
Figure 1.6 Newton’s differential calculus
Turning to integration, we note that ideas similar to “line segments, surfaces,
solids [being] made up of an actual infinite number of indivisible or infinitesimal
elements” were employed by Kepler, Cavalieri, and other predecessors of Newton and
Leibniz in “their determinations of areas and volumes” (Ehrlich [2006] 494). Newton’s
first treatment of integration contained the following set of three rules
Rule 1: If y ax m / n , then the area under y is an /n m x m / n 1 . Rule 2: If y is given by the sum of more terms (also an infinite number of terms),
y y1 y2 ..., then the area under y is given by the sum of the areas of the corresponding terms. Rule 3: In order to calculate the area under a curve f (x,y) 0, one must expand y as a sum of terms of the form ax m / n and apply Rule 1 and Rule 2. (Guicciardini [2003] 76).
This version of ‘termwise integration’ was motivated by Newton’s heavy use of binomial series. Newton justified rule 1 by proving the fundamental theorem of the calculus
(hereafter the FTC; formally presented in section I.B). His proof that the integral of the
derivative of z is equal to z rests on the “commonplace” assumption that the area under a
6 Taken from Guicciardini [2003] p. 79 14 cuve was considered equal to the sum of infinitely many infinitesimal strips.7 Going the
other way—i.e. proving that the derivative of the integral of z is equal to z—Newton’s
argument is outlined by Guicciardini as follows.
Newton considered a curve AD (see Fig. 3.2), where AB x, BD y and the area ABD z. He defined B o and BK v such that “the rectangle BHK( ov) is equal to the space BD.”…With these definitions one has that A x o and the area A is equal to z ov. At this point, Newton wrote: “from any arbitrarily assumed relationship x and z I seek y.” He noted that the increment of the area ov, divided by the increment of the abscissa o is equal to v. But since one can assume “ B to be infinitely small, that is, o to be zero, v and y will be equal.” Therefore, the rate of increase of the area is equal to the ordinate. (Guicciardini [2003] 77).