The True History of The Limit

Fermat, Leibniz, Euler, and the
Gang: The True History of the

Concepts of Limit and Shadow
Tiziana Bascelli, Emanuele Bottazzi, Frederik Herzberg,
Vladimir Kanovei, Karin U. Katz, Mikhail G. Katz,
Tahl Nowik, David Sherry, and Steven Shnider
T
he theories as developed by European Fermat determined the optimal value by impos-
mathematicians prior to 1870 differed ing a condition using his adequality of quantities.
from the modern ones in that none of But he did not really think of quantities as func-
them used the modern theory of limits. tions, nor did he realize that his method produced
Fermat develops what is sometimes only a necessary condition for his optimization
called a “precalculus” theory, where the optimal condition. For a more detailed general introduc-
value is determined by some special condition such tion, see chapters 1 and 2 of the volume edited by
as equality of roots of some equation. The same Grattan-Guinness (Bos et al. 1980 [19]).
can be said for his contemporaries like Descartes, The doctrine of limits is sometimes claimed to
Huygens, and Roberval. have replaced that of infinitesimals when analysis
Leibniz’s calculus advanced beyond them in was rigorized in the nineteenth century. While
working on the derivative function of the variable x. it is true that Cantor, Dedekind and Weierstrass
He had the indefinite integral whereas his prede- attempted (not altogether successfully; see Ehrlich
cessors only had concepts more or less equivalent 2006 [32], Mormann & Katz 2013 [79]) to eliminate
to it. Euler, following Leibniz, also worked with infinitesimals from analysis, the history of the limit
such functions, but distinguished the variable (or concept is more complex. Newton had explicitly
variables) with constant differentials dx, a status written that his ultimate ratios were not actually
that corresponds to the modern assignment that x ratios but, rather, limits of prime ratios (see Russell
is the independent variable, the other variables of 1903 [89, item 316, pp. 338-339]; Pourciau 2001
[84]). In fact, the sources of a rigorous notion of
the problem being dependent upon it (or them)
limit are considerably older than the nineteenthth
functionally.
century.
Tiziana Bascelli is an independent researcher in history In the context of Leibnizian mathematics, the
and philosophy of science. Her email address is tiziana. limit of f (x) as x tends to x0 can be viewed as
bascelli@virgilio.it. the “assignable part” (as Leibniz may have put
Emanuele Bottazzi is a Ph.D. student at the Università di it) of f (x0 + dx) where dx is an “inassignable”
Trento, Italy. His email address is Emanuele.Bottazzi@ infinitesimal increment (whenever the answer is
unitn.it. independent of the infinitesimal chosen). A modern
Frederik Herzberg is an assistant professor of mathemati- formalization of this idea exploits the standard
cal economics at Bielefeld University, Germany, as well as part principle (see Keisler 2012 [67, p. 36]).
an external member of the Munich Center for Mathematical
Philosophy, Germany. His email address is fherzberg@uni- Mikhail G. Katz is professor of mathematics at Bar Ilan Uni-
bielefeld.de. versity, Israel. His email is katzmik@macs.biu.ac.il.
Vladimir Kanovei is professor of mathematics at IPPI, Tahl Nowik is professor of mathematics at Bar Ilan Univer-
Moscow, and MIIT, Moscow, Russia. His email address is sity, Israel. His email is tahl@math.biu.ac.il.
kanovei@googlemail.com. David Sherry is professor of philosophy at Northern Arizona
Karin U. Katz teaches mathematics at Bar Ilan University, University. His email address is David.Sherry@nau.edu.
Israel. Her email address is katzmik@macs.biu.ac.il. Steven Shnider is professor of mathematics at Bar Ilan
DOI: http://dx.doi.org/10.1090/noti1149 University, Israel. His email is shnider@macs.biu.ac.il.
848 Notices of the AMS Volume 61, Number 8

In the context of ordered fields E, the standard −1 0 1 2 3 4
part principle is the idea that, if E is a proper
extension of the real numbers R, then every finite
(or limited) element x ∈ E is infinitely close to
a suitable x0 ∈ R. Such a real number is called
the standard part (sometimes called the shadow)
of x, or in formulas, st(x) = x0 . Denoting by Ef the
r +α r r +β r +γ
collection of finite elements of E, we obtain a map
st : Ef → R.
Here x is called finite if it is smaller (in absolute
value) than some real number (the term finite is
immediately comprehensible to a wide mathemati-
cal public, whereas limited corresponds to correct st
technical usage); an infinitesimal is smaller (in

absolute value) than every positive real; and x is −1 0 1 r 2 3 4
infinitely close to x0 in the sense that x − x0 is
infinitesimal. Figure 1. The standard part function, st, “rounds
Briefly, the standard part function “rounds off” off” a finite hyperreal to the nearest real number.
a finite element of E to the nearest real number The function st is here represented by a vertical
(see Figure 1). projection. An “infinitesimal microscope” is used
to view an infinitesimal neighborhood of a
The proof of the principle is easy. A finite
standard real number r , where α , β , and γ
element x ∈ E defines a Dedekind cut on the
represent typical infinitesimals. Courtesy of
subfield R ⊂ E (alternatively, on Q ⊂ R), and
Wikipedia.
the cut in turn defines the real x0 via the usual
correspondence between cuts and real numbers.
One sometimes writes down the relation
Methodological Remarks
x ≈ x0
To comment on the historical subtleties of judging
to express infinite closeness. or interpreting past mathematics by present-day
We argue that the sources of such a relation, standards,1 note that neither Fermat, Leibniz, Euler,
and of the standard part principle, go back to nor Cauchy had access to the semantic founda-
Fermat, Leibniz, Euler, and Cauchy. Leibniz would tional frameworks as developed in mathematics
discard the inassignable part of 2x + dx to arrive at the end of the nineteenthth and first half of
at the expected answer, 2x, relying on his law of
the twentieth centuries. What we argue is that
homogeneity (see the section entitled “Leibniz’s
their syntactic inferential moves ultimately found
Transcendental Law of Homogeneity”). Such an
modern proxies in Robinson’s framework, thus
inferential move is mirrored by a suitable proxy in
placing a firm (relative to ZFC)2 semantic foun-
the hyperreal approach, namely the standard part
dation underneath the classical procedures of
function.
these masters. Benacerraf (1965 [10]) formulated
Fermat, Leibniz, Euler, and Cauchy all used
a related dichotomy in terms of mathematical
one or another form of approximate equality, or
the idea of discarding “negligible” terms. Their practice vs. mathematical ontology.
inferential moves find suitable proxies in the For example, the Leibnizian laws of continuity
context of modern theories of infinitesimals, and (see Knobloch 2002 [69, p. 67]) and homogene-
specifically the concept of shadow. ity can be recast in terms of modern concepts
The last two sections present an application of such as the transfer principle and the standard
the standard part to decreasing rearrangements part principle over the hyperreals, without ever
of real functions and to a problem on divergent appealing to the semantic content of the technical
integrals due to S. Konyagin. development of the hyperreals as a punctiform con-
This article continues efforts in revisiting the tinuum; similarly, Leibniz’s proof of the product
history and foundations of infinitesimal calculus rule for differentiation is essentially identical, at
and modern nonstandard analysis. Previous efforts the syntactic level, to a modern infinitesimal proof
in this direction include Bair et al. (2013 [6]), (see, again, the section “Leibniz’s Transcendental
Bascelli (2014 [7]), Błaszczyk et al. (2013 [15]), Law of Homogeneity”).
Borovik et al. (2012 [16], [17]), Kanovei et al. (2013
[55]), Katz, Katz & Kudryk (2014 [61]), Mormann 1
Some reflections on this can be found in Lewis (1975 [76]).
et al. (2013 [79]), Sherry et al. (2014 [92]), Tall et 2
The Zermelo–Fraenkel Set Theory with the Axiom of
al. (2014 [97]). Choice.
September 2014 Notices of the AMS 849

A-track and B-track formulated a condition, in terms of the mean value
The crucial distinction between syntactic and se- theorem,5 for what would qualify as a successful
mantic aspects of the work involving mathematical theory of infinitesimals, and concluded:
continua appears to have been overlooked by I will not say that progress in this direction
R. Arthur who finds fault with the hyperreal proxy is impossible, but it is true that none of
of the Leibnizian continuum, by arguing that the the investigators have achieved anything
latter was non-punctiform (see Arthur 2013 [5]). positive (Klein 1908 [68, p. 219]).
Yet this makes little difference at the syntactic Klein was referring to the current work on
level, as explained above. Arthur’s brand of the syn- infinitesimal-enriched systems by Levi-Civita, Bet-
categorematic approach following Ishiguro (1990 tazzi, Stolz, and others. In Klein’s mind, the
[52]) involves a reductive reading of Leibnizian infinitesimal track was very much a current re-
infinitesimals as logical (as opposed to pure) fic- search topic; see Ehrlich (2006 [32]) for a detailed
tions involving a hidden quantifier à la Weierstrass, coverage of the work on infinitesimals around 1900.
ranging over “ordinary” values. This approach was
critically analyzed in (Katz & Sherry 2013 [65]), Formal Epistemology: Easwaran on Hyperreals
(Sherry & Katz 2013 [92]), and (Tho 2012 [101]). Some recent articles are more encouraging in that
Robinson’s framework poses a challenge to they attempt a more technically sophisticated ap-
traditional historiography of mathematical analysis. proach. K. Easwaran’s study (2014 [31]), motivated
The traditional thinking is often dominated by a by a problem in formal epistemology,6 attempts to
kind of Weierstrassian teleology. This is a view deal with technical aspects of Robinson’s theory
of the history of analysis as univocal evolution such as the notion of internal set, and shows an
toward the radiant Archimedean framework as awareness of recent technical developments, such
developed by Cantor, Dedekind, Weierstrass, and as a definable hyperreal system of Kanovei & Shelah
others starting around 1870, described as the (2004 [57]).
A-track in a recent piece in the Notices (see Bair et Even though Easwaran, in the tradition of
al. 2013 [6]). Lewis (1980 [77]) and Skyrms (1980 [94]), tries to
Robinson’s challenge is to point out not only engage seriously with the intricacies of employing
the possibility, but also the existence of a parallel hyperreals in formal epistemology,7 not all of his
Bernoullian3 track for the development of analysis, findings are convincing. For example, he assumes
or B-track for short. The B-track assigns an that physical quantities cannot take hyperreal
irreducible and central role to the concept of values.8 However, there exist physical quantities
infinitesimal, a role it played in the work of Leibniz, that are not directly observable. Theoretical proxies
Euler, mature Lagrange,4 Cauchy, and others. for unobservable physical quantities typically
The caliber of some of the response to Robin- depend on the chosen mathematical model. And,
son’s challenge has been disappointing. Thus, the not surprisingly, there are mathematical models
critique by Earman (1975 [30]) is marred by a of physical phenomena which operate with the
confusion of second-order infinitesimals like dx2 hyperreals, in which physical quantities take
and second-order hyperreal extensions like ∗∗ R; hyperreal values. Many such models are discussed
see (Katz & Sherry 2013 [65]) for a discussion. in the volume by Albeverio et al. (1986 [1]).
Victor J. Katz (2014 [66]) appears to imply For example, certain probabilistic laws of nature
that a B-track approach based on notions of have been formulated using hyperreal-valued prob-
infinitesimals or indivisibles is limited to “the work ability theory. The construction of mathematical
of Fermat, Newton, Leibniz, and many others in
5
the seventeenth and eighteenth centuries.” This The Klein–Fraenkel criterion is discussed in more detail in
does not appear to be Felix Klein’s view. Klein Kanovei et al. (2013 [55]).
6
The problem is concerned with saving philosophical
3 Bayesianism, a popular position in formal epistemology,
Historians often name Johann Bernoulli as the first mathe-
matician to have adhered systematically and exclusively to which appears to require that one be able to find on every
the infinitesimal approach as the basis for the calculus. algebra of doxastically relevant propositions some subjec-
4 tive probability assignment such that only the impossible
In the second edition of his Mécanique Analytique dating
from 1811, Lagrange fully embraced the infinitesimal in the event (∅) will be assigned an initial/uninformed subjective
following terms: “Once one has duly captured the spirit of probability, or credence, of 0.
7
this system [i.e., infinitesimal calculus], and has convinced For instance, he concedes: “And the hyperreals may also
oneself of the correctness of its results by means of the geo- help, as long as we understand that they do not tell us
metric method of the prime and ultimate ratios, or by means the precise structure of credences.” (Easwaran 2014 [31],
of the analytic method of derivatives, one can then exploit Introduction, last paragraph).
8
the infinitely small as a reliable and convenient tool so as Easwaran’s explicit premise is that “All physical quanti-
to shorten and simplify proofs.” See (Katz & Katz 2011 [58]) ties can be entirely parametrized using the standard real
for a discussion. numbers.” (Easwaran 2014 [31, Section 8.4, Premise 3]).

Brownian motion by Anderson (1976 [4]) provides
a hyperreal model of the botanical counterpart. It
is unclear why (and indeed rather implausible that)
an observer A, whose degrees of belief about botan-
ical Brownian motion stem from a mathematical
model based on the construction of mathematical
Brownian motion by Wiener (1923 [104]), should
be viewed as being more rational than another
observer B, whose degrees of belief about botanical
Brownian motion stem from a mathematical model
based on Anderson’s construction of mathematical Figure 2. Easwaran’s attempted slaying of the
infinitesimal, following P. Uccello. Uccello’s
Brownian motion.9
creature is shown as inhabiting an infinitesimal
Similarly problematic is Easwaran’s assumption
neighborhood of 0 .
that an infinite sequence of probabilistic tests must
necessarily be modeled by the set of standard
natural numbers (this is discussed in more detail in
the subsection “Williamson, Complexity, and Other that the hyperreals do not exist (Easwaran 2014
Arguments”). Such an assumption eliminates the [30, Section 8.4]); see Figure 2. However, on the
possibility of modeling it by a sequence of infinite same grounds, one would have to reject parts of
hypernatural length. Indeed, once one allows for mathematics with important applications. There
infinite sequences to be modeled in this way, the are fundamental results in functional analysis
problem of assigning a probability to an infinite that depend on the Axiom of Choice such as the
sequence of coin tosses that was studied in (Elga Hahn–Banach theorem; yet no one would suggest
2004 [33]) and (Williamson 2007 [105]) allows for that mathematical physicists or mathematical
an elegant hyperreal solution (Herzberg 2007 [48]). economists should stop exploiting them.
Easwaran reiterates the common objection that Most real analysis textbooks prove the σ -
the hyperreals are allegedly “nonconstructive” additivity (i.e., countable additivity) of Lebesgue
entities. The bitter roots of such an allegation in the measure, but σ -additivity is not deducible from
radical constructivist views of E. Bishop have been ZF, as shown by the Feferman–Levy model; see
critically analyzed in (Katz & Katz 2011 [59]), and (Feferman & Levy 1963 [36]); (Jech 1973 [54,
contrasted with the liberal views of the intuitionist chapter 10]). Indeed, it is consistent with ZF that
A. Heyting, who felt that Robinson’s theory was the following holds:
“a standard model of important mathematical
(∗) the continuum R ofS real numbers is a
research” (Heyting 1973 [51, p. 136]). It is important
countable union R = n∈N Xn of countable
to keep in mind that Bishop’s target was classical
sets Xn .
mathematics (as a whole), the demise of which he
predicted in the following terms: See (Cohen 1966 [26, chapter IV, section 4]) for a
Very possibly classical mathematics will description of a model of ZF in which (∗) holds.10
cease to exist as an independent discipline Note that (∗) implies that the Lebesgue measure
(Bishop 1968 [14, p. 54]). is not countably additive, as all countable sets are
null sets whereas R is not a null set. Therefore
countable additivity of the Lebesgue measure
Zermelo–Fraenkel Axioms and the Feferman–
Levy Model cannot be established in ZF.
Terence Tao wrote:
In his analysis, Easwaran assigns substantial weight
to the fact that “it is consistent with the ZF [Zermelo– By giving up countable additivity, one loses
Fraenkel set theory] without the Axiom of Choice” a fair amount of measure and integration
theory, and in particular the notion of the
9
One paradoxical aspect of Easwaran’s methodology is that, expectation of a random variable becomes
despite his anti-hyperreal stance in (2014 [30]), he does en- problematic (unless the random variable
vision the possibility of useful infinitesimals in an earlier takes only finitely many values). (Tao 2013
joint paper (Colyvan & Easwaran 2008 [27]), where he cites
[100])
John Bell’s account (Bell’s presentation of Smooth Infinitesi-
mal Analysis in [9] involves a category-theoretic framework
10
based on intuitionistic logic); but never the hyperreals. Fur- Property (∗) may appear to be asserting the countabil-
thermore, in the 2014 paper he cites the surreals as possible ity of the continuum. However, in order to obtain a bijective
alternatives to the real number–based description of the map from a countable collection of countable sets to N × N
“structure of physical space” as he calls it; see subsection (and hence, by diagonalization, to N), the Axiom of Choice
“Williamson, Complexity, and Other Arguments” for a more (in its “countable” version which allows a countably-infinite
detailed discussion. sequence of independent choices) will necessarily be used.

Tao’s remarks suggest that deducibility from ZF Viewing a purely constructive Skolem hypernatural
is not a reasonable criterion of mathematical
H ∈ NSko \ N
plausibility by any modern standard.
There are models of ZF in which there are as a member of NRob via the inclusion (1), one can
infinitesimal numbers, if properly understood, apply the transfer principle to form the set
among the real numbers themselves. Thus, there XH = {A ⊂ N : H ∈ ∗A},
exist models of ZF which are also models of Nelson’s
(1987 [82]) radically elementary mathematics, a where ∗A ⊂ NRob is the natural extension of A.
subsystem of Nelson’s (1977 [81]) Internal Set The set XH is not measurable. What propels the
Theory. Here radically elementary mathematics is set XH ⊂ P(N) into existence is not a purported
an extension of classical set theory (which may be weakness of a nonstandard integer H itself, but
understood as ZF11 ) by a unary predicate, to be rather the remarkable strength of both the Łoś-
interpreted as Robinson transfer principle and the consequences
it yields.
“… is a standard natural number,”
with additional axioms that regulate the use of Williamson, Complexity, and Other Arguments
the new predicate (notably external induction
Easwaran makes a number of further critiques of
for standard natural numbers) and ensure the
hyperreal methodology. His section 8.1, entitled
existence of nonstandard numbers. Nelson (1987
“Williamson’s Argument,” concerns infinite coin
[82, Appendix]) showed that a major part of the
tosses. Easwaran’s analysis is based on the model
theory of continuous-time stochastic processes
of a countable sequence of coin tosses given by
is in fact equivalent to a corresponding radically
Williamson [105]. In this model, it is assumed that
elementary theory involving infinitesimals, and
indeed, radically elementary probability theory has … for definiteness, [the coin] will be flipped
seen applications in the sciences; see for example once per second, assuming that seconds
(Reder 2003 [85]). from now into the future can be numbered
In sum, mathematical descriptions of nontrivial with the natural numbers (Easwaran 2014
natural phenomena involve, by necessity, some [31, section 8.1]).
degree of mathematical idealization, but Easwaran What is lurking behind this is a double assumption
has not given us a good reason why only such which, unlike other “premises,” is not made explicit
mathematical idealizations that are feasible in by Easwaran. Namely, he assumes that
every model of ZF should be acceptable. Rather, (1) a vast number of independent tests is best
as we have already seen, there are very good modeled by a temporal arrangement thereof,
arguments (e.g., from measure theory) against rather than by a simultaneous collection;
such a high reverence for ZF. and
(2) the collection of seconds ticking away “from
Skolem Integers and Robinson Integers now [and] into the future” gives a faithful
Easwaran recycles the well-known claim by representation of the natural numbers.
A. Connes that a hypernatural number leads to These two premises are not self-evident and
a nonmeasurable set. However, the criticism by some research mathematicians have very different
Connes12 is in the category of dressing down a intuitions about the matter, as much of the
feature to look like a bug, to reverse a known literature on applied nonstandard analysis (e.g.,
dictum from computer science slang.13 This can be Albeverio et al. 1986 [1], Reder 2003 [85]) illustrates.
seen as follows. The Skolem nonstandard integers It seems that in Easwaran’s model, an agent can
NSko are known to be purely constructive; see choose not to flip the coin at some seconds, thus
Skolem (1955 [93]) and Kanovei et al. (2013 [55]). giving rise to events like “a coin that is flipped
Yet they imbed in Robinson’s hypernaturals NRob : starting at second 2 comes up heads on every flip.”
(1) NSko > NRob . However, in all applications we are aware of, this
additional structure used to rule out the use of
11
Even though Nelson would probably argue for a much hyperreals as the range of probability functions
weaker system; see Herzberg (2013 [49, Appendix A.1]), seems not to be relevant.
citing Nelson (2011 [83]).
12
Williamson and Easwaran appear to be unwilling
Note that Connes relied on the Hahn-Banach theorem,
to assume that, once one decides to use hyperreal
exploited ultrafilters, and placed a nonconstructive entity
infinitesimals, one should also replace the original
(namely the Dixmier trace) on the front cover of his mag-
num opus; see (Katz & Leichtnam 2013 [62]) and (Kanovei algebra “of propositions in which the agent has
et al. 2013 [55]) for details. credence” with an internal algebra of the hyperreal
13 setting. In fact, such an additional step allows one
See https://en.wikipedia.org/wiki/Undocumented_
feature to avoid both the problems raised by Williamson’s

argument in his formulation using conditional measures. However, Easwaran is apparently com-
probability, and those raised by Easwaran in fortable with the idealization of exploiting a larger
section 8.2 of his paper. number system than the rationals for the value
A possible model with hyperreal infinitesi- range of probability measures. What we argue
mals for an infinite sequence of coin tosses is is that the real numbers are merely one among
given by representing every event by means of possible idealizations that can be used for this
a sequence {a1 , . . . , aN }, where an represents the purpose. For instance, in hyperreal models for
outcome of the nth flip and N is a fixed hypernat- infinite sequence of coin tosses developed by Benci,
ural number. In this model, consider the events Bottazzi, & Di Nasso (2013 [11]), all events have
“an = Heads for n ≤ N”, which we will denote H(1), hyperrational probabilities. This generalizes both
and “an = Heads for 2 ≤ n ≤ N”, that we will the case of finite sequences of coin tosses, and
denote H(2). In such a setting, events H(1) and the Kolmogorovian model for infinite sequences
H(2) are not isomorphic, contrary to what was of coin tosses, where a real-valued probability is
argued in (Williamson [105, p. 3]). This is due to the generated by applying Carathéodory’s extension
fact that hypernatural numbers are an elementary theorem to the rational-valued probability measure
extension of the natural numbers, for which the over the cylinder sets.
formula k 6= k + 1 always holds. Moreover, the Given Easwaran’s firm belief that “the function
probability of H(1) is the infinitesimal 2−N , while relating credences to the physical is not so complex
the probability of H(2) is the strictly greater that its existence is independent of Zermelo-
infinitesimal 2−(N−1) , thus obeying the well-known Fraenkel set theory” (see his section 8.4, premise
rule for conditional probability. 2), it is surprising to find him suggesting that
Easwaran’s section 8.4 entitled “The complexity
the surreal numbers seem more promising
argument” is based on four premises. However, his
as a device for future philosophers of
premise 3, to the effect that “all physical quantities
probability to use (Easwaran 2014 [31,
can be entirely parameterized using the standard
Appendix A.3]).
real numbers,” is unlikely to lead to meaningful
philosophical conclusions based on “first princi- However, while the construction of the surre-
ples.” This is because all physical quantities can als indeed “is a simultaneous generalization of
be entirely parameterized by the usual rational Dedekind’s construction of the real numbers and
numbers alone, due to the intrinsic limits of our von Neumann’s construction of the ordinals,” as
capability to measure physical quantities. A clear observed by Easwaran, it is usually carried out
explanation of this limitation was given by Dowek. in the Von Neumann–Bernays–Gödel set theory
In particular, since (NBG) with Global Choice; see, for instance, the
“Preliminaries” section of (Alling 1987 [3]). The
a measuring instrument yields only an ap-
assumption of the Global Axiom of Choice is a
proximation of the measured magnitude,
strong foundational assumption.
[…] it is therefore impossible, except accord-
The construction of the surreal numbers can
ing to this idealization, to measure more
be performed within a version of NBG that is a
than the first digits of a physical magnitude.
conservative extension of ZFC, but does not need
[…] According to this principle, this ideal-
ization of the process of measurement is a Limitation of Size (or Global Choice). However, NBG
fiction. This suggests the idea, reminiscent clearly is not a conservative extension of ZF; and if
of Pythagoras’ views, that Physics could be one wishes to prove certain interesting features of
formulated with rational numbers only. We the surreals one needs an even stronger version
can therefore wonder why real numbers of NBG that involves the Axiom of Global Choice.
have been invented and, moreover, used in Therefore the axiomatic foundation that one needs
Physics. A hypothesis is that the invention for using the surreal numbers is at least as strong
of real numbers is one of the many situa- as the one needed for the hyperreals.
tions, where the complexity of an object is
increased, so that it can be apprehended Infinity and Infinitesimal: Let Both Pretty
more easily. (Dowek 2013 [29]) Severely Alone
Related comments by Wheeler (1994 [103, p. 308]), At the previous turn of the century, H. Heaton
Brukner & Zeilinger (2005 [22, p. 59]), and oth- wrote:
ers were analyzed by Kanovei et al. (2013 [55, I think I know exactly what is meant by the
Section 8.4]). See also Jaroszkiewicz (2014 [53]). term zero. But I can have no conception
If all physical quantities can be entirely parame- either of infinity or of the infinitesimal, and
terized by using rational numbers, there should I think it would be well if mathematicians
be no compelling reason to choose the real num- would let both pretty severely alone (Heaton
ber system as the value range of our probability 1898 [47, p. 225]).

Heaton’s sentiment expresses an unease about a (6) Among the remaining terms, suppress all
mathematical concept of which one may have an terms which still contain a factor of e.
intuitive grasp14 but which is not easily formal- Solving the resulting equation for a yields
izable. Heaton points out several mathematical the extremum of f .
inconsistencies or ill-chosen terminology among In modern mathematical language, the algorithm
the conceptions of infinitesimals of his contempo- entails expanding the difference quotient
raries. This highlights the brilliant mathematical f (a + e) − f (a)
achievement of a consistent “calculus” for infinites- e
imals attained through the work of Hewitt (1948
in powers of e and taking the constant term.15
[50]), Łoś (1955 [78]), Robinson (1961 [87]), and
The method (leaving aside step (5)) is immediately
Nelson (1977 [81]), but also of their predecessors understandable to a modern reader as the elemen-
like Fermat, Euler, Leibniz, and Cauchy, as we tary calculus exercise of finding the extremum
analyze respectively in sections entitled “Fermat’s by solving the equation f 0 (a) = 0. But the real
Adequality,” “Leibniz’s Transcendental Law of Ho- question is how Fermat understood this algorithm
mogeneity,” “Euler’s Principle of Cancellation,” and in his own terms, in the mathematical language
“What Did Cauchy Mean by Limit?”. of his time, prior to the invention of calculus by
Barrow, Leibniz, Newton, and others.
Fermat’s Adequality There are two crucial points in trying to under-
Our interpretation of Fermat’s technique is compat- stand Fermat’s reasoning: first, the meaning of
ible with those by Strømholm (1968 [95]) and Giusti “adequality” in step (2), and second, the justifica-
(2009 [43]). It is at variance with the interpretation tion for suppressing the terms involving positive
by Breger (1994 [21]), considered by Knobloch powers of e in step (6). The two issues are closely
(2014 [70]) to have been refuted. related because interpretation of adequality de-
Adequality, or π αρισ óτης (parisotēs) in the pends on the conditions on e. One condition which
original Greek of Diophantus, is a crucial step Fermat always assumes is that e is positive. He did
in Fermat’s method of finding maxima, minima, not use negative numbers in his calculations.16
Fermat introduces the term adequality in Metho-
tangents, and solving other problems that a modern
dus with a reference to Diophantus of Alexandria.
mathematician would solve using infinitesimal
In the third article of the series, Ad Eamdem
calculus. The method is presented in a series of
Methodum (Sur la Même Méthode), he quotes
short articles in Fermat’s collected works. The first
Diophantus’s Greek term π αρισ óτης, which
article, Methodus ad Disquirendam Maximam et he renders following Xylander and Bachet, as
Minimam, opens with a summary of an algorithm adaequatio or adaequalitas (see A. Weil [102,
for finding the maximum or minimum value p. 28]).
of an algebraic expression in a variable A. For
convenience, we will write such an expression in Tangent Line and Convexity of Parabola
modern functional notation as f (a).
Consider Fermat’s calculation of the tangent line
to the parabola (see Fermat [38, pp. 122–123]). To
Summary of Fermat’s Algorithm simplify Fermat’s notation, we will work with the
One version of the algorithm can be broken up parabola y = x2 , or
into six steps in the following way: x2
(1) Introduce an auxiliary symbol e, and = 1.
y
form f (a + e);
15
(2) Set adequal the two expressions f (a +e) =AD Fermat also envisions a more general technique involving
f (a) (the notation “=AD ” for adequality is division by a higher power of e as in step (5).
16
ours, not Fermat’s); This point is crucial for our argument below using the
transverse ray. Since Fermat is only working with positive
(3) Cancel the common terms on the two sides
values of his e, he only considers a ray (rather than a full
of the adequality. The remaining terms all line) starting at a point of the curve. The convexity of the
contain a factor of e; curve implies an inequality, which Fermat transforms into
(4) Divide by e (see also next step); an adequality without giving much explanation of his pro-
(5) In a parenthetical comment, Fermat adds: cedure, but assuming implicitly that the ray is tangent to
“or by the highest common factor of e;” the curve. But a transverse ray would satisfy the inequality
no less than a tangent ray, indicating that Fermat is relying
on an additional piece of geometric information. His proce-
14
The intuitive appeal of infinitesimals make them an dure of applying the defining relation of the curve itself to a
effective teaching tool. The pedagogical value of teach- point on the tangent ray is only meaningful when the incre-
ing calculus with infinitesimals was demonstrated in a ment e is small (see subsection “Tangent Line and Convexity
controlled study by Sullivan (1976 [96]). of Parabola”).

To understand what Fermat is doing, it is helpful never considered negative roots, and if A =
to think of the parabola as a level curve of the 0 was a solution of an equation, he did not
2
two-variable function xy . mention it as it was nearly always geomet-
Given a point (x, y) on the parabola, Fermat rically uninteresting (Strømholm 1968 [95,
wishes to find the tangent line through the p. 49]).
point. Fermat exploits the geometric fact that, Fermat says nothing about considering
by convexity, a point points y + e “on the other side,” i.e., further away
from the vertex of the parabola, as he does in the
(p, q) context of applying a related but different method,
on the tangent line lies outside the parabola. He for instance in his two letters to Mersenne (see [95,
therefore obtains an inequality equivalent in our p. 51]), and in his letter to Brûlart [39].18 Now for
p2
notation to q > 1, or p2 > q. Here q = y − e, and e positive values of e, Fermat’s inequality (2) would
be satisfied by a transverse ray (i.e., secant ray)
is Fermat’s magic symbol we wish to understand.
starting at (x, y) and lying outside the parabola,
Thus we obtain
just as much as it is satisfied by a tangent ray
p2 starting at (x, y). Fermat’s method therefore
(2) > 1.
y −e presupposes an additional piece of information,
At this point Fermat proceeds as follows: privileging the tangent ray over transverse rays.
p2 The additional piece of information is geometric
(i) he writes down the inequality y−e > 1, in origin: he applies the defining relation (of the
or p2 > y − e; curve itself) to a point on the tangent ray to the
(ii) he invites the reader to adégaler (to “ade- curve, a procedure that is only meaningful when
quate”); the increment e is small.
2 y
(iii) he writes down the adequality px 2 =AD y−e ; In modern terms, we would speak of the tangent
(iv) he uses an identity involving similar triangles line being a “best approximation” to the curve
to substitute for a small variation e; however, Fermat does not
x y +r explicitly discuss the size of e. The procedure of
= “discarding the remaining terms” in step (v) admits
p y +r −e
of a proxy in the hyperreal context. Namely, it is
where r is the distance from the vertex of the standard part principle (see the Introduction).
the parabola to the point of intersection of Fermat does not elaborate on the justification of
the tangent to the parabola at y with the axis this step, but he is always careful to speak of the
of symmetry, suppressing or deleting the remaining term in e,
(v) he cross multiplies and cancels identical rather than setting it equal to zero. Perhaps his
terms on right and left, then divides out by e, rationale for suppressing terms in e consists in
discards the remaining terms containing e, ignoring terms that don’t correspond to an actual
and obtains y = r as the solution.17 measurement, prefiguring Leibniz’s inassignable
What interests us here are steps (i) and (ii). quantities. Fermat’s inferential moves in the context
How does Fermat pass from an inequality to an of his adequality are akin to Leibniz’s in the context
adequality? Giusti noted that of his calculus; see the section called “Leibniz’s
Transcendental Law of Homogeneity”.
Comme d’habitude, Fermat est autant dé-
taillé dans les exemples qu’il est réticent
Fermat, Galileo, and Wallis
dans les explications. On ne trouvera donc
presque jamais des justifications de sa règle While Fermat never spoke of his e as being infinitely
des tangentes (Giusti 2009 [43]). small, the technique was known both to Fermat’s
contemporaries like Galileo (see Bascelli 2014 [7],
In fact, Fermat provides no explicit explanation
[8]) and Wallis (see Katz & Katz [60, Section 24])
for this step. However, what he does is apply as well as Fermat himself, as his correspondence
the defining relation for a curve to points on with Wallis makes clear; see Katz, Schaps & Shnider
the tangent line to the curve. Note that here the (2013 [63, Section 2.1]).
quantity e, as in q = y − e, is positive: Fermat did Fermat was very interested in Galileo’s treatise
not have the facility we do of assigning negative De motu locali, as we know from his letters to Marin
values to variables. Strømholm notes that Fermat Mersenne dated Apr/May 1637, 10 August, and 22
October 1638. Galileo’s treatment of infinitesimals
17
In Fermat’s notation y = d, y + r = a. Step (v) can be un- in De motu locali is discussed by Wisan (1974 [106,
y (y+r )2
derstood as requiring the expression y−e − (y+r −e)2 to have p. 292]) and Settle (1966 [91]).
a double root at e = 0, leading to the solution y = r or in
18
Fermat’s notation a = 2r . This was noted by Giusti (2009 [43]).

Alexander (2014 [2]) notes that the clerics in Il faut donc adégaler (à cause de la propriété
Rome forbade the doctrine of the infinitely small spécifique de la courbe qui est à considérer
on 10 August 1632 (a month before Galileo was put sur la tangente)
on trial over heliocentrism); this may help explain (see Katz et al. (2013 [63]) for more details).
why the Catholic Fermat might have been reluctant Fermat’s approach involves applying the defining
to speak of the infinitely small explicitly.19 relation of the curve, to a point on a tangent to the
In a recent text, U. Felgner analyzes the Dio- curve. The approach is consistent with the idea of
phantus problems which exploit the method of approximation inherent in his method, involving a
π αρισ óτης, and concludes that negligible distance (whether infinitesimal or not)
Aus diesen Beispielen wird deutlich, dass between the tangent and the original curve when
die Verben π άρισ oυ̃ν und adaequare nicht one is near the point of tangency. This line of
ganz dasselbe ausdrücken. Das griechische reasoning is related to the ideas of the differential
Wort bedeutet, der Gleichheit nahe zu sein, calculus. Note that Fermat does not say anything
während das lateinische Wort das Erreichen- here concerning the multiplicities of zeros of
der Gleichheit (sowohl als vollendeten als polynomials. As Felgner himself points out, in the
auch als unvollendeten Prozeß) ausdrückt case of the cycloid the only polynomial in sight is
(Felgner 2014 [37]). of first order and the increment “e” cancels out.
Thus, in his view, even though the two expres- Fermat correctly solves the problem by obtaining
sions have slightly different meanings, the Greek the defining equation of the tangent.
meaning “being close to equality” and the Latin For a recent study of seventeenth century
meaning “equality which is reached (at the end methodology, see the article (Carroll et al. 2013
of either a finite or an infinite process),” they [23]).
both involve approximation. Felgner goes on to
consider some of the relevant texts from Fermat, Leibniz’s Transcendental Law of
and concludes that Fermat’s method has nothing Homogeneity
to do with differential calculus and involves only In this section, we examine a possible connection
the property of an auxiliary expression having a between Fermat’s adequality and Leibniz’s Tran-
double zero: scendental Law of Homogeneity (TLH). Both of
them enable certain inferential moves that play
Wir hoffen, deutlich gemacht zu haben,
parallel roles in Fermat’s and Leibniz’s approaches
dass die fermatsche “Methode der Adaequa-
to the problem of maxima and minima. Note the
tio” gar nichts mit dem Differential-Kalkül
similarity in titles of their seminal texts: Methodus
zu hat, sondern vielmehr im Studium des
ad Disquirendam Maximam et Minimam (Fermat,
Wertverlaufs eines Polynoms in der Umge-
see Tannery [98, pp. 133]) and Nova methodus
bung eines kritischen Punktes besteht, und
pro maximis et minimis … (Leibniz 1684 [72] in
dabei das Ziel verfolgt zu zeigen, dass
Gerhardt [42]).
das Polynom an dieser Stelle eine doppelte
Nullstelle besitzt. (ibid.)
When Are Quantities Equal?
However, Felgner’s conclusion is inconsistent with
Leibniz developed the TLH in order to enable
his own textual analysis which indicates that the
inferences to be made between inassignable and
idea of approximation is present in the methods of
assignable quantities. The TLH governs equations
both Diophantus and Fermat. As Knobloch (2014
involving differentials. H. Bos interprets it as
[70]) notes, “Fermat’s method of adequality is not
follows:
a single method but rather a cluster of methods.”
Felgner failed to analyze the examples of tangents A quantity which is infinitely small with
to transcendental curves, such as the cycloid, in respect to another quantity can be neglected
which Fermat does not study the order of the if compared with that quantity. Thus all
zero of an auxiliary polynomial. Felgner mistakenly terms in an equation except those of the
asserts that, in the case of the cycloid, Fermat highest order of infinity, or the lowest order
did not reveal how he thought of the solution: of infinite smallness, can be discarded. For
“Wie FERMATsich die Lösung dachte, hat er nicht instance,
verraten.” (ibid.) Quite to the contrary, as Fermat (3) a + dx = a
explicitly stated, he applied the defining property
dx + ddy = dx
of the curve to points on the tangent line:
etc. The resulting equations satisfy this
19
See a related discussion at http://math. […] requirement of homogeneity (Bos 1974
stackexchange.com/questions/661999/are- [18, p. 33] paraphrasing Leibniz 1710 [75,
infinitesimals-dangerous. pp. 381–382]).

The title of Leibniz’s 1710 text is Symbolismus where a “is any finite quantity” (see Euler 1755
memorabilis calculi algebraici et infinitesimalis in [35, § § 86,87]) are consonant with a Leibnizian
comparatione potentiarum et differentiarum, et de tradition as reported by Bos; see formula (3) above.
lege homogeneorum transcendentali. The inclusion To explain formulas like (5), Euler elaborated
of the transcendental law of homogeneity (lex two distinct ways (arithmetic and geometric) of
homogeneorum transcendentalis) in the title of the comparing quantities, in the following terms:
text attests to the importance Leibniz attached to Since we are going to show that an infinitely
this law. small quantity is really zero, we must
The “equality up to an infinitesimal” implied in meet the objection of why we do not
TLH was explicitly discussed by Leibniz in a 1695 always use the same symbol 0 for infinitely
response to Nieuwentijt, in the following terms: small quantities, rather than some special
Caeterum aequalia esse puto, non tantum ones…[S]ince we have two ways to compare
quorum differentia est omnino nulla, sed them, either arithmetic or geometric, let us
et quorum differentia est incomparabiliter look at the quotients of quantities to be
parva; et licet ea Nihil omnino dici non de- compared in order to see the difference.
beat, non tamen est quantitas comparabilis If we accept the notation used in the
cum ipsis, quorum est differentia (Leibniz analysis of the infinite, then dx indicates
1695 [73, p. 322]) [emphasis added–authors] a quantity that is infinitely small, so that
We provide a translation of Leibniz’s Latin: both dx = 0 and a dx = 0, where a is any
Besides, I consider to be equal not only finite quantity. Despite this, the geomet-
those things whose difference is entirely ric ratio a dx : dx is finite, namely a : 1.
nothing, but also those whose difference is For this reason, these two infinitely small
incomparably small: and granted that it [i.e., quantities, dx and a dx, both being equal
the difference] should not be called entirely to 0, cannot be confused when we consider
Nothing, nevertheless it is not a quantity their ratio. In a similar way, we will deal
comparable to those whose difference it is. with infinitely small quantities dx and dy
(ibid., § 86, pp. 51–52) [emphasis added–the
Product Rule authors].
How did Leibniz use the TLH in developing the Having defined the arithmetic and geometric com-
calculus? The issue can be illustrated by Leibniz’s parisons, Euler proceeds to clarify the difference
justification of the last step in the following between them as follows:
calculation: Let a be a finite quantity and let dx be
d(uv) = (u + du)(v + dv) − uv infinitely small. The arithmetic ratio of
equals is clear: Since ndx = 0, we have
(4) = udv + vdu + du dv
a ± ndx − a = 0.
= udv + vdu.
On the other hand, the geometric ratio is
The last step in the calculation (4) depends on the
clearly of equals, since
following inference:
a ± ndx
d(uv) = udv +vdu+dudv =⇒ d(uv) = udv +vdu. (6) = 1.
a
Such an inference is an application of Leibniz’s TLH. From this we obtain the well-known rule that
In his 1701 text Cum Prodiisset [74, pp. 46–47], the infinitely small vanishes in comparison
Leibniz presents an alternative justification of the with the finite and hence can be neglected
product rule (see Bos [18, p. 58]). Here he divides [with respect to it] [35, §87] [emphasis in
by dx, and argues with differential quotients rather the original–the authors].
than differentials. The role played by the TLH
Like Leibniz, Euler considers more than one way
in these calculations is similar to that played
of comparing quantities. Euler’s formula (6) indi-
by adequality in Fermat’s work on maxima and
cates that his geometric comparison is procedurally
minima. For more details on Leibniz, see Guillaume
identical with the Leibnizian TLH.
(2014 [45]); Katz & Sherry (2012 [64]), (2013 [65]);
To summarize, Euler’s geometric comparison
Sherry & Katz [92]; Tho (2012 [101]).
of a pair of quantities amounts to their ratio
being infinitely close to a finite quantity, as in
Euler’s Principle of Cancellation
formula (6); the same is true for TLH. Note
Some of the Leibnizian formulas reappear, not that one has a + dx = a in this sense for an
surprisingly, in his student’s student Euler. Euler’s appreciable a 6= 0, but not for a = 0 (in which
formulas like case there is equality only in the arithmetic sense).
(5) a + dx = a, Euler’s “geometric” comparison was dubbed “the

principle of cancellation” in (Ferraro [40, pp. 47, accroissements petits des variables. Leibniz
48, 54]). et les premiers géomètres qui se sont oc-
Euler proceeds to present the usual rules of cupés de l’analyse infinitésimale ont appelé
infinitesimal calculus, which go back to Leibniz, différentielles des variables leurs accroisse-
l’Hôpital, and the Bernoullis, such as ments infiniment petits, … (Cauchy 1844
[25, p. 5]).
(7) a dxm + b dxn = a dxm
Two important points emerge from this passage.
provided m < n “since dxn vanishes compared First, Cauchy specifically speaks about neglecting
with dxm ” ([35, § 89]), relying on his “geometric” (“on néglige”) higher-order terms, rather than
comparison. Euler introduces a distinction between setting them equal to zero. This indicates a
infinitesimals of different order, and directly similarity of procedure with the Leibnizian TLH
computes 20 a ratio of the form (see the section “Leibniz’s Transcendental Law of
dx ± dx2 Homogeneity”). Like Leibniz and Fermat before
= 1 ± dx = 1
dx him, Cauchy does not set the higher-order terms
of two particular infinitesimals, assigning the equal to zero, but rather “neglects” or discards
value 1 to it (ibid., § 88). Euler concludes: them. Furthermore, Cauchy’s comments on Leibniz
deserve special attention.
Although all of them [infinitely small quan-
tities] are equal to 0, still they must be
Cauchy on Leibniz
carefully distinguished one from the other
if we are to pay attention to their mutual By speaking matter-of-factly about the infinitesi-
relationships, which has been explained mals of Leibniz specifically, Cauchy reveals that
through a geometric ratio (ibid., § 89). his (Cauchy’s) infinitesimals are consonant with
Leibniz’s. This is unlike the differentials where
The Eulerian hierarchy of orders of infinitesimals
Cauchy adopts a different approach.
harks back to Leibniz’s work (see the section
On page 6 of the same text, Cauchy notes that
“Leibniz’s Transcendental Law of Homogeneity”).
the notion of derivative
Euler’s geometric comparison, or “principle of
cancellation,” is yet another incarnation of the idea représente en réalité la limite du rapport
at the root of Fermat’s adequality and Leibniz’s entre les accrossements infiniment petits et
Transcendental Law of Homogeneity. For further simultanés de la fonction et de la variable
details on Euler see Bibiloni et al. (2006 [13]); Bair (ibid., p. 6) [emphasis added–the authors]
et al. (2013 [6]); Reeder (2013 [86]). The same definition of the derivative is repeated on
page 7, this time emphasized by means of italics.
What Did Cauchy Mean by “Limit”? Note Cauchy’s emphasis on the point that the
Laugwitz’s detailed study of Cauchy’s methodology derivative is not a ratio of infinitesimal increments,
places it squarely in the B-track (see the section but rather the limit of the ratio.
called “Methodological Remarks”). In conclusion, Cauchy’s use of the term “limit” as applied
Laugwitz writes: to a ratio of infinitesimals in this context may
be unfamiliar to a modern reader, accustomed
The influence of Euler should not be ne-
to taking limits of sequences of real numbers.
glected, with regard both to the organization
Its meaning is clarified by Cauchy’s discussion
of Cauchy’s texts and, in particular, to the
of “neglecting” higher order infinitesimals in the
fundamental role of infinitesimals (Laugwitz
previous paragraph on page 5 cited above. Cauchy’s
1987 [71, p. 273]).
use of “limit” is procedurally identical with the
Thus, in his 1844 text Exercices d’analyse et de Leibnizian TLH, and therefore similarly finds its
physique mathématique, Cauchy wrote: modern proxy as extracting the standard part out
…si, les accroissements des variables étant of the ratio of infinitesimals.
supposés infiniment petits, on néglige, On page 11, Cauchy chooses infinitesimal
vis-à-vis de ces accroissements consid- increments ∆s and ∆t, and writes down the
érés comme infiniment petits du premier equation
ordre, les infiniment petits des ordres ds
(8) = lim. .
∆s
supérieurs au premier, les nouvelles équa- dt ∆t
tions deviendront linéaires par rapport aux
Modulo replacing Cauchy’s symbol “lim.” by the
20 modern one “st” or “sh,” Cauchy’s formula (8) is
Note that Euler does not “prove that the expression
is equal to 1;” such indirect proofs are a trademark of
identical to the formula appearing in any textbook
the (, δ) approach. Rather, Euler directly computes (what based on the hyperreal approach, expressing the
would today be formalized as the standard part of) the derivative in terms of the standard part function
expression. (shadow).

Cauchy on Continuity Modern Formalizations: A Case Study
On page 17 of his 1844 text, Cauchy gives a To illustrate the use of the standard part in the
definition of continuity in terms of infinitesimals context of the hyperreal field extension of R, we
(an infinitesimal x-increment necessarily produces will consider the following problem on divergent
an infinitesimal y-increment). His definition is integrals. The problem was recently posed at
nearly identical with the italicized definition that SE, and is reportedly due to S. Konyagin.21 The
appeared on page 34 in his Cours d’Analyse (Cauchy solution exploits the technique of a monotone
1821 [24]) 23 years earlier, when he first introduced rearrangement g of a function f , shown by Ryff
the modern notion of continuity. We will use the to admit a measure-preserving map φ : [0, 1] →
translation by Bradley & Sandifer (2009 [20]). In his [0, 1] such that f = g ◦ φ. In general there is
Section 2.2 entitled Continuity of functions, Cauchy no “inverse” ψ such that g = f ◦ ψ; however, a
writes: hyperreal enlargement enables one to construct a
suitable (internal) proxy for such a ψ, so as to be
If, beginning with a value of x contained
able to write g = st(f ◦ ψ); see formula (14) below.
between these limits, we add to the variable x
an infinitely small increment α, the function Theorem 1. Let f be a real-valued function contin-
itself is incremented by the difference f (x + uous on [0, 1]. Then there exists a number a such
α) − f (x). that the integral
Z1
Cauchy goes on to state that 1
(10) dx
the function f (x) is a continuous function 0 |f (x) − a|
of x between the assigned limits if, for diverges.
each value of x between these limits, the
numerical value of the difference f (x + A proof can be given in terms of a monotone
α) − f (x) decreases indefinitely with the rearrangement of the function (see Hardy et
numerical value of α. al. [46]). We take a decreasing rearrangement g(x)
of the function f (x). If f is continuous, then the
He then proceeds to provide an italicized definition
function g(x) will also be continuous. If f is not
of continuity in the following terms:
constant on any set of positive measure, one can
the function f (x) is continuous with respect construct g by setting
to x between the given limits if, between (11)
these limits, an infinitely small increment g = m−1 where m(y) = meas{x : f (x) > y}.
in the variable always produces an infinitely
Ryff (1970 [90]) showed that there exists a
small increment in the function itself.
measure-preserving transformation22 φ : [0, 1] →
In modern notation, Cauchy’s definition can be [0, 1] that relates f and g as follows:

stated as follows. Denote by x the halo of x, i.e.,

the collection of all points infinitely close to x. (12) f (x) = g ◦ φ(x).
Then f is continuous at x if Finding a map ψ such that g(x) = f ◦ ψ(x) is

in general impossible (see Bennett & Sharpley
(9) f x ⊂ f (x) . [12, p. 85, example 7.7] for a counterexample). This
Most scholars hold that Cauchy never worked with difficulty can be circumvented using a hyperfinite
a pointwise definition of continuity (as is customary rearrangement (see the section entitled “A Combi-
today) but rather required a condition of type (9) natorial Approach to Decreasing Rearrangements”).
to hold in a range (“between the given limits”). It By measure preservation, we have
Z1 Z1
is worth recalling that Cauchy never gave an , δ
definition of either limit or continuity (though |f (x) − a|−1 dx = |g(x) − a|−1 dx
0 0
(, δ)-type arguments occasionally do appear in
(for every a).23
Cauchy). It is a widespread and deeply rooted
To complete the proof of Theorem 1, ap-
misconception among both mathematicians and
ply the result that every monotone function is
those interested in the history and philosophy of
mathematics that it was Cauchy who invented the 21
http://math.stackexchange.com/questions/408311/
modern (, δ) definitions of limit and continuity; improper-integral-diverges
see, e.g., Colyvan & Easwaran (2008 [27, p. 88]) 22
However, see the section “A Combinatorial Approach to
who err in attributing the formal (, δ) definition Decreasing Rearrangements” for a hyperfinite approach
of continuity to Cauchy. That this is not the case avoiding measure theory altogether.
23 −1 by
was argued by Błaszczyk et al. (2013 [15]), Borovik Here one needs to replace the function |f (x)− a|
−1
et al. (2012 [17]), Katz & Katz (2011 [58]), Nakane the family of its truncations min C, |f (x) − a| , and then
(2014 [80]), Tall et al. (2014 [97]). let C increase without bound.

a.e. differentiable.24 Take a point p ∈ [0, 1] where laws as the usual, “standard” domain of real
the function g is differentiable. Then the num- numbers and related objects. Thus, as for finite
ber a = g(p) yields an infinite integral (10), since sets, there exists a permutation ψ of the hyperfinite
the difference |g(x) − a| can be bounded above in grid
terms of a linear expression.25 (13) GH = {p1 , . . . , pH }
by decreasing value of f (pi ) (here f (ψ(p1 )) is
A Combinatorial Approach to Decreasing
the maximal value). We assume that equal values
Rearrangements are ordered lexicographically so that if f (pi ) =
The existence of a decreasing rearrangement of a f (pj ) with i < j then ψ(pi ) < ψ(pj ). Hence we
function f continuous on [0, 1] admits an elegant obtain an internal function
proof in the context of its hyperreal extension ∗f ,
(14) ĝ(pi ) = f (ψ(pi )), i = 1, . . . , H.
which we will continue to denote by f .
We present a combinatorial argument showing Here ĝ is (perhaps nonstrictly) decreasing on the
that the decreasing rearrangement enjoys the same grid GH of (13). The internal sequences (f (pi ))
modulus of uniformity as the original function.26 and (ĝ(pi )), where i = 1, . . . , H, are equinumerable
The argument actually yields an independent con- in the sense above.
struction of the decreasing rearrangement (see Proposition 2. Let f be an arbitrary continuous
Proposition 2) that avoids recourse to measure function. Then there is a standard continuous real
theory. It also yields an “inverse up to an infini- function g(x) such that g(st(pi )) = st(ĝ(pi )) for
tesimal,” ψ (see formula (14)), to the function φ all i, where st(y) denotes the standard part of a
such that f = g ◦ φ. For a recent application of hyperreal y.
combinatorial arguments in a hyperreal framework,
Proof. Let gi = ĝ(pi ). We claim that ĝ is S-
see Benci et al. (2013 [11]).
continuous (microcontinuous), i.e., for each
In passing from the finite to the continuous
pair i, j = 1, . . . , H, if pi − pj is infinitesimal then
case of rearrangements, Bennett and Sharpley [12]
so is ĝ(pi ) − ĝ(pj ). To prove the claim, we will
note that
prove the following stronger fact:
nonnegative sequences (a1 , a2 , . . . , an ) for every i < j there are m < n such that n−
and (b1 , b2 , . . . , bn ) are equimeasurable m ≤ j − i and |f (pm ) − f (pn )| ≥ gi − gj .
if and only if there is a permutation
The sets A = {k : f (pk ) ≥ gi } and B = {k : f (pk ) ≤
σ of {1, 2, . . . , n} such that bi = aσ (i)
gj } are nonempty and there are at most j − i − 1
for i = 1, 2, . . . , n. … The notion of permu- points which are not in A ∪ B. Let m ∈ A and n ∈
tation is no longer available in this context B be such that |m − n| is minimal. All integers
[of continuous measure spaces] and is between m and n are not in A∪B. Hence there are at
replaced by that of a “measure-preserving most j −i −1 such integers, and therefore |n−m| ≤
transformation” (Bennett and Sharpley j −i. By the definition of A and B, we obtain |f (pn )−
1988 [12, p. 79]). f (pm )| ≥ gi − gj , which proves the claim. Thus ĝ
We show that the hyperreal framework allows one is indeed S-continuous.
to continue working with combinatorial ideas, such This allows us to define, for any standard x ∈
as the “inverse” function ψ, in the continuous case [0, 1], the value g(x) to be the standard part of the
as well. hyperreal gi for any hyperinteger i such that pi is
i
Let H ∈ ∗ N \ N, let pi = H for i = 1, 2, . . . , H. By infinitely close to x, and then g is a continuous27
the Transfer Principle (see e.g., Davis [28], Herzberg and (nonstrictly) monotone real function equal to
[49], Kanovei & Reeken [56]), the nonstandard the decreasing rearrangement g = m−1 of (11).
domain of internal sets satisfies the same basic
The hyperreal approach makes it possible to
24
In fact, one does not really need to use the result that
solve Konyagin’s problem without resorting to
monotone functions are a.e. differentiable. Consider the standard treatments of decreasing rearrangements
convex hull in the plane of the graph of the monotone func- which use measure theory. Note that the rearrange-
tion g(x), and take a point where the graph touches the ment defined by the internal permutation ψ
boundary of the convex hull (other than the endpoints 0 preserves the integral of f (as well as the integrals
and 1). Setting a equal to the y-coordinate of the point does of the truncations of |f (x) − a|−1 ), in the following
the job. sense. The right-hand Riemann sums satisfy
25
Namely, for x near such a point p, we have |g(x) − a| ≤
1 1 H H H
(|g 0 (p)| + 1)|x − p|, hence |g(x)−a| ≥ (|g 0 (p)|+1)|x−a| ,
X X X
(15) f (pi )∆x = f (ψ(pi ))∆x = ĝ(pi )∆x,
yielding a lower bound in terms of a divergent integral. i=1 i=1 i=1
26
A function f on [0, 1] is said to satisfy a modulus of
uniformity µ(n) > 0, n ∈ N, if ∀n ∈ 27
N ∀p, q ∈ The argument shows in fact that the modulus of unifor-
1
[0, 1] |p − q| ≤ µ(n) → |f (p) − f (q)| ≤ n . mity of g is bounded by that of f ; see footnote 26.

where ∆x = H1 . Thus ψ transforms a hyperfinite and S. Shnider, Is mathematical history written
Riemann sum of f into a hyperfinite Riemann by the victors? Notices of the American Math-
ematical Society 60 (2013) no. 7, 886–904. See
R1 P
H
sum of ĝ. Since 0 f (x)dx = st i=1 f (pi )∆x
http://www.ams.org/notices/201307/rnoti-
and g(st(pi )) = st ĝ(pi ) , we conclude that f p886.pdf and http://arxiv.org/abs/1306.5973.
and g have the same integrals, and similarly for [7] T. Bascelli, Galileo’s quanti: Understanding infini-
the integrals of |f (x) − a|−1 ; see footnote 23. tesimal magnitudes, Arch. Hist. Exact Sci. 68 (2014),
The first equality in (15) holds automatically no. 2, 121–136.
[8] , Infinitesimal issues in Galileo’s theory of mo-
by the transfer principle even though ψ is an
tion, Revue Roumaine de Philosophie 58 (2014), no. 1,
infinite permutation. (Compare with the standard 23-41.
situation where changing the order of summation [9] J. Bell, The Continuous and the Infinitesimal in
in an infinite sum generally requires further Mathematics and Philosophy, Polimetrica, 2006.
justification.) This illustrates one of the advantages [10] P. Benacerraf, What numbers could not be, Philos.
of the hyperreal approach. Rev. 74 (1965), 47–73.
[11] V. Benci, E. Bottazzi, and M. Di Nasso, Elementary
numerosity and measures, preprint (2013).
Conclusion [12] C. Bennett and R. Sharpley, Interpolation of op-
We have critically reviewed several common mis- erators, Pure and Applied Mathematics vol. 129,
representations of hyperreal number systems, not Academic Press, Boston, MA, 1988.
least in relation to their alleged nonconstruc- [13] L. Bibiloni, P. Viader, and J. Paradís, On a series of
tiveness, from a historical, philosophical, and Goldbach and Euler, Amer. Math. Monthly 113 (2006),
no. 3, 206–220.
set-theoretic perspective. In particular we have
[14] E. Bishop, Mathematics as a numerical language,
countered some of Easwaran’s recent arguments 1970 Intuitionism and Proof Theory (Proc. Conf.,
against the use of hyperreals in formal epistemol- Buffalo, N.Y., 1968), pp. 53–71, North-Holland,
ogy. A hyperreal framework enables a richer syntax Amsterdam.
better suited for expressing proxies for procedural [15] P. Błaszczyk, M. Katz, and D. Sherry, Ten
moves found in the work of Fermat, Leibniz, Euler, misconceptions from the history of anal-
ysis and their debunking, Foundations of
and Cauchy. Such a framework sheds light on
Science 18 (2013), no. 1, 43–74. See http://
the internal coherence of their procedures which dx.doi.org/10.1007/s10699-012-9285-8 and
have been often misunderstood from a whiggish http://arxiv.org/abs/1202.4153.
post-Weierstrassian perspective. [16] A. Borovik, R. Jin, and M. Katz, An integer
construction of infinitesimals: Toward a the-
Acknowledgments ory of Eudoxus hyperreals, Notre Dame Journal
of Formal Logic 53 (2012), no. 4, 557-570. See
The work of Vladimir Kanovei was partially sup- http://dx.doi.org/10.1215/00294527-1722755
ported by RFBR grant 13-01-00006. M. Katz was and http://arxiv.org/abs/1210.7475.
partially funded by the Israel Science Foundation [17] A. Borovik and M. Katz, Who gave you the
grant no. 1517/12. We are grateful to Thomas Cauchy–Weierstrass tale? The dual history
Mormann and to the anonymous referee for help- of rigorous calculus, Foundations of Sci-
ful suggestions, and to Ivor Grattan-Guinness for ence 17 (2012), no. 3, 245–276. See http://
dx.doi.org/10.1007/s10699-011-9235-x and
contributing parts of the introduction.
http://arxiv.org/abs/1108.2885.
[18] H. J. M. Bos, Differentials, higher-order differentials
References and the derivative in the Leibnizian calculus, Arch.
[1] S. Albeverio, R. Høegh-Krohn, J. Fenstad, and History Exact Sci. 14 (1974), 1–90.
T. Lindstrøm, Nonstandard methods in stochastic [19] H. J. M. Bos, R. Bunn, J. Dauben, I. Grattan-
analysis and mathematical physics, Pure and Applied Guinness, T. Hawkins, and K. M. Pedersen, From
Mathematics 122, Academic Press, Inc., Orlando, FL, the calculus to set theory, 1630–1910, An introduc-
1986. tory history, Edited by I. Grattan-Guinness, Gerald
[2] A. Alexander, Infinitesimal: How a Dangerous Math- Duckworth & Co. Ltd., London, 1980.
ematical Theory Shaped the Modern World, Farrar, [20] R. Bradley and C. Sandifer, Cauchy’s Cours
Straus and Giroux, 2014. d’analyse, An annotated translation, Sources and
[3] N. Alling, Foundations of Analysis over Surreal Studies in the History of Mathematics and Physical
Number Fields, North-Holland Mathematical Library, Sciences, Springer, New York, 2009.
1987. [21] H. Breger, The mysteries of adaequare: A vindication
[4] R. Anderson, A nonstandard representation for of Fermat, Arch. Hist. Exact Sci. 46 (1994), no. 3, 193–
Brownian motion and Itô integration, Israel Journal 219.
of Mathematics 25 (1976), no. 1-2, 15–46. [22] Č. Brukner and A. Zeilinger, Quantum physics as
[5] R. Arthur, Leibniz’s syncategorematic infinitesimals, a science of information, in Quo vadis quantum me-
Arch. Hist. Exact Sci. 67 (2013), no. 5, 553–593. chanics?, 47–61, Frontiers Collection, Springer, Berlin,
[6] J. Bair, P. Błaszczyk, R. Ely, V. Henry, 2005.
V. Kanovei, K. Katz, M. Katz, S. Kutate- [23] M. Carroll, S. Dougherty, and D. Perkins, Indivis-
ladze, T. McGaffey, D. Schaps, D. Sherry, ibles, infinitesimals and a tale of seventeenth-century

mathematics, Mathematics Magazine 86 (2013), no. 4, [44] U. Goldenbaum and D. Jesseph (eds.), Infinitesimal
239–254. Differences: Controversies between Leibniz and his
[24] A. L. Cauchy, Cours d’Analyse de L’Ecole Contemporaries, Berlin-New York: Walter de Gruyter,
Royale Polytechnique, Première Partie, Analyse 2008.
algébrique, Paris: Imprimérie Royale, 1821, On- [45] M. Guillaume, “Review of M. Katz, D. Sherry,
line at http://books.google.com/books?id= Leibniz’s infinitesimals: Their fictionality, their
_mYVAAAAQAAJ&dq=cauchy&lr=&source=gbs_ modern implementations, and their foes from
navlinks_s. Berkeley to Russell and beyond, Erkenntnis 78 (2013),
[25] , Exercices d’analyse et de physique mathéma- no. 3, 571–625.” Mathematical Reviews (2014). See
tique, (vol. 3), Paris, Bachelier, 1844. http://www.ams.org/mathscinet-getitem?mr=
[26] P. Cohen, Set Theory and the Continuum Hypothesis 3053644.
W. A. Benjamin, New York-Amsterdam, 1966. [46] G. Hardy, J. Littlewood, and G. Pólya, Inequalities,
[27] M. Colyvan and K. Easwaran, Mathematical and Second edition, Cambridge University Press, 1952.
physical continuity, Australas. J. Log. 6 (2008), 87–93. [47] H. Heaton, Infinity, the infinitesimal, and zero,
[28] M. Davis, Applied nonstandard analysis, Pure American Mathematical Monthly 5 (1898), no. 10,
and Applied Mathematics, Wiley-Interscience 224–226.
[John Wiley & Sons], New York-London-Sydney, [48] F. Herzberg, Internal laws of probability, general-
1977. Reprinted: Dover, NY, 2005, see http:// ized likelihoods and Lewis’ infinitesimal chances—a
store.doverpublications.com/0486442292. response to Adam Elga. British Journal for the
html. Philosophy of Science 58 (2007), no. 1, 25–43.
[29] G. Dowek, Real numbers, chaos, and the prin- [49] , Stochastic calculus with infinitesimals, Lecture
ciple of a bounded density of information, Notes in Mathematics, 2067, Springer, Heidelberg,
Invited paper at International Computer Sci- 2013.
ence Symposium in Russia, 2013. See http:// [50] E. Hewitt, Rings of real-valued continuous functions,
who.rocq.inria.fr/Gilles.Dowek/Publi/csr. I, Trans. Amer. Math. Soc. 64 (1948), 45–99.
pdf. [51] A. Heijting, Address to Professor A. Robinson, At
[30] J. Earman, Infinities, infinitesimals, and indivisibles: the occasion of the Brouwer memorial lecture given
The Leibnizian labyrinth, Studia Leibnitiana 7 (1975), by Professor A. Robinson on April 26, 1973, Nieuw
no. 2, 236–251. Archief voor Wiskunde (3) 21 (1973), 134–137.
[31] K. Easwaran, Regularity and hyperreal credences, [52] H. Ishiguro, Leibniz’s Philosophy of Logic and Lan-
Philosophical Review 123 (2014), No. 1, 1–41. guage, Second edition, Cambridge University Press,
[32] P. Ehrlich, The rise of non-Archimedean mathemat- Cambridge, 1990.
ics and the roots of a misconception, I, The emergence [53] G. Jaroszkiewicz, Principles of Discrete Time Me-
of non-Archimedean systems of magnitudes, Arch. chanics, Cambridge Monographs on Mathematical
Hist. Exact Sci. 60 (2006), no. 1, 1–121. Physics, Cambridge University Press, 2014.
[33] A. Elga, Infinitesimal chances and the laws of nature, [54] T. Jech, The axiom of choice, Studies in Logic
Australasian Journal of Philosophy 82 (2004), no. 1, and the Foundations of Mathematics, Vol. 75.
67–76. North-Holland Publishing Co., Amsterdam-London;
[34] L. Euler, Institutiones Calculi Differentialis, SPb, American Elsevier Publishing Co., Inc., New York,
1755. 1973.
[35] , Foundations of Differential Calculus, English [55] V. Kanovei, M. Katz, and T. Mormann, Tools,
translation of Chapters 1–9 of ([34]) by D. Blanton, objects, and chimeras: Connes on the role of hyper-
Springer, New York, 2000. reals in mathematics, Foundations of Science
[36] S. Feferman and A. Levy, Independence results in 18 (2013), no. 2, 259–296. See http://
set theory by Cohen’s method II, Notices Amer. Math. dx.doi.org/10.1007/s10699-012-9316-5 and
Soc. 10 (1963), 593. http://arxiv.org/abs/1211.0244.
[37] U. Felgner, Der Begriff der “Angleichung” [56] V. Kanovei and M. Reeken, Nonstandard analysis,
(π αρισ óτης, adaequatio) bei Diophant und axiomatically, Springer Monographs in Mathematics,
Fermat (2014), preprint. Berlin: Springer, 2004.
[38] P. Fermat, Méthode pour la recherche du maximum [57] V. Kanovei and S. Shelah, A definable nonstandard
et du minimum, pp. 121–156 in Tannery’s edition model of the reals, Journal of Symbolic Logic 69
[99]. (2004), no. 1, 159–164.
[39] , Letter to Brûlart, Oeuvres, Vol. 5, pp. 120–125. [58] K. Katz and M. Katz, Cauchy’s continuum,
[40] G. Ferraro, Differentials and differential co- Perspectives on Science 19 (2011), no. 4, 426–
efficients in the Eulerian foundations of the 452. See http://arxiv.org/abs/1108.4201 and
calculus, Historia Mathematica 31 (2004), no. 1, http://www.mitpressjournals.org/doi/abs/
34–61. See http://dx.doi.org/10.1016/S0315- 10.1162/POSC_ a_00047.
0860(03)00030-2. [59] , Meaning in classical mathematics: Is it at odds
[41] C. I. Gerhardt (ed.), Historia et Origo calculi differ- with intuitionism? Intellectica 56 (2011), no. 2, 223–
entialis a G. G. Leibnitio conscripta, ed. C. I. Gerhardt, 302. See http://arxiv.org/abs/1110.5456.
Hannover, 1846. [60] , A Burgessian critique of nominalistic ten-
[42] , Leibnizens mathematische Schriften, (Berlin dencies in contemporary mathematics and its
and Halle: Eidmann, 1850–1863). historiography, Foundations of Science 17 (2012),
[43] E. Giusti, Les méthodes des maxima et minima de no. 1, 51–89. See http://dx.doi.org/10.1007/
Fermat, Ann. Fac. Sci. Toulouse Math. (6) 18 (2009), s10699-011-9223-1 and http://arxiv.org/abs/
Fascicule Special, 59–85. 1104.0375.

[61] K. Katz, M. Katz, and T. Kudryk, Toward a clar- tialem siu infinitesimalem motas, in Gerhardt [42], V,
ity of the extreme value theorem, Logica Uni- p. 320–328, 1695.
versalis 8 (2014), no. 2, 193-214. See http:// [74] , Cum Prodiisset …mss “Cum prodiisset atque
dx.doi.org/10.1007/s11787-014-0102-8 and increbuisset Analysis mea infinitesimalis…” in
http://arxiv.org/abs/1404.5658. Gerhardt [41], pp. 39–50, 1701.
[62] M. Katz and E. Leichtnam, Commuting and non- [75] , Symbolismus memorabilis calculi algebraici
commuting infinitesimals, American Mathematical et infinitesimalis in comparatione potentiarum et
Monthly 120 (2013), no. 7, 631–641. See http:// differentiarum, et de lege homogeneorum tran-
dx.doi.org/10.4169/amer.math.monthly.120. scendentali, in Gerhardt [42, vol. V, pp. 377–382],
07.631 and http://arxiv.org/abs/1304.0583. 1710.
[63] M. Katz, D. Schaps, and S. Shnider, Almost [76] A. Lewis, Bringing modern mathematics to bear on
equal: The method of adequality from Dio- historical research, Historia Mathematica 2 (1975),
phantus to Fermat and beyond. Perspectives 84–85.
on Science 21 (2013), no. 3, 283–324. See [77] , A subjectivist’s guide to objective chance.
http://www.mitpressjournals.org/doi/abs/10. In Studies in Inductive Logic and Probability (ed.
1162/POSC_a_00101 and http://arxiv.org/abs/ Richard C. Jeffrey), pp. 263–293, University of
1210.7750. California Press, 1980.
[64] M. Katz and D. Sherry, Leibniz’s laws of conti- [78] J. Łoś, Quelques remarques, théorèmes et problèmes
nuity and homogeneity, Notices of the American sur les classes définissables d’algèbres, in Mathe-
Mathematical Society 59 (2012), no. 11, 1550–1558. matical interpretation of formal systems, 98–113,
See http://www.ams.org/notices/201211/ and North-Holland Publishing Co., Amsterdam, 1955.
http://arxiv.org/abs/1211.7188. [79] T. Mormann and M. Katz, Infinitesimals as an issue
[65] , Leibniz’s infinitesimals: Their fictional- of neo-Kantian philosophy of science, HOPOS: The
ity, their modern implementations, and their Journal of the International Society for the History of
foes from Berkeley to Russell and beyond, Er- Philosophy of Science 3 (2013), no. 2, 236–280. See
kenntnis 78 (2013), no. 3, 571–625. See http:// http://www.jstor.org/stable/10.1086/671348.
dx.doi.org/10.1007/s10670-012-9370-y and and http://arxiv.org/abs/1304.1027.
http://arxiv.org/abs/1205.0174. [80] M. Nakane, Did Weierstrass’s differential cal-
[66] V. Katz, “Review of Bair et al., Is math- culus have a limit-avoiding character? His
ematical history written by the victors? definition of a limit in -δ style, BSHM Bulletin:
Notices Amer. Math. Soc. 60 (2013), no. 7, Journal of the British Society for the History
886–904.” Mathematical Reviews (2014). See of Mathematics 29 (2014), no. 1, 51–59. See
http://www.ams.org/mathscinet-getitem?mr= http://dx.doi.org/10.1080/17498430.2013.
3086638. 831241.
[67] H. J. Keisler, Elementary Calculus: An Infinitesimal [81] E. Nelson, Internal set theory: a new approach
Approach, Second Edition, Prindle, Weber & Schimdt, to nonstandard analysis. Bulletin of the American
Boston, 1986. Revision from February 2012 online Mathematical Society 83 (1977), no. 6, 1165–1198.
at http://www.math.wisc.edu/~keisler/calc. [82] , Radically elementary probability theory, An-
html. nals of Mathematics Studies, vol. 117, Princeton
[68] F. Klein, Elementary Mathematics from an Advanced University Press, 1987.
Standpoint, Vol. I, Arithmetic, Algebra, Analysis, [83] , Warning signs of a possible collapse of con-
Translation by E. R. Hedrick and C. A. Noble [Macmil- temporary mathematics, in Infinity (Michael Heller
lan, New York, 1932] from the third German edition and W. Hugh Woodin, eds.), pp. 76–85, Cambridge
[Springer, Berlin, 1924]. Originally published as Ele- University Press, 2011.
mentarmathematik vom höheren Standpunkte aus [84] B. Pourciau, Newton and the notion of limit, Historia
(Leipzig, 1908). Mathematica 28 (2001), no. 1, 18–30.
[69] E. Knobloch, Leibniz’s rigorous foundation of infin- [85] C. Reder, Transient behaviour of a Galton-Watson
itesimal geometry by means of Riemannian sums, process with a large number of types, Journal of
Foundations of the formal sciences, 1 (Berlin, 1999), Applied Probability 40 (2003), no. 4, 1007–1030.
Synthese 133 (2002), no. 1–2, 59–73. [86] P. Reeder, Internal Set Theory and Euler’s Introduc-
[70] , “Review of: M. Katz, D. Schaps, and tio in Analysin Infinitorum, MSc Thesis, Ohio State
S. Shnider, Almost equal: The method of University, 2013.
adequality from Diophantus to Fermat and be- [87] A. Robinson, Nonstandard analysis, Nederl. Akad.
yond, Perspectives on Science 21 (2013), no. 3, Wetensch. Proc. Ser. A 64, Indag. Math. 23 (1961),
283–324.” Mathematical Reviews (2014). See 432–440 [reprinted in Selected Works, see item [88],
http://www.ams.org/mathscinet-getitem?mr= pp. 3–11]
3114421. [88] , Selected papers of Abraham Robinson, Vol. II,
[71] D. Laugwitz, Infinitely small quantities in Cauchy’s Nonstandard analysis and philosophy, edited and
textbooks, Historia Mathematica 14 (1987), no. 3, with introductions by W. A. J. Luxemburg and S.
258–274. Körner, Yale University Press, New Haven, Conn.,
[72] G. Leibniz, Nova methodus pro maximis et minimis 1979.
…, in Acta Erud., October 1684. See Gerhardt [42], V, [89] B. Russell, The Principles of Mathematics, Routledge,
pp. 220–226. London 1903.
[73] , Responsio ad nonnullas difficultates a [90] J. Ryff, Measure preserving transformations and
Dn. Bernardo Niewentiit circa methodum differen- rearrangements, J. Math. Anal. Appl. 31 (1970),
449–458.

MATHE M ATI CS AT T H E N AT I O N AL S EC U RI T Y A G ENCY [91] T. Settle, Galilean science: Essays in the mechan-
8 ics and dynamics of the Discorsi, Ph.D. dissertation,
Rise Above the Ordinary

Cornell University, 1966, 288 pages.
[92] D. Sherry and M. Katz, Infinitesimals,
imaginaries, ideals, and fictions, Studia
Leibnitiana 44 (2012), no. 2, 166–192. See
http://arxiv.org/abs/1304.2137.
[93] T. Skolem, Peano’s axioms and models of arithmetic,
in Mathematical interpretation of formal systems,
pp. 1–14. North-Holland Publishing Co., Amsterdam,
1955.
[94] B. Skyrms, Causal Necessity: A Pragmatic Investiga-
tion of the Necessity of Laws, Yale University Press,
1980.
[95] P. Strømholm, Fermat’s methods of maxima and min-
ima and of tangents, A reconstruction, Arch. History
Exact Sci. 5 (1968), no. 1, 47–69.
[96] K. Sullivan, Mathematical Education: The Teaching
of Elementary Calculus Using the Nonstandard Anal-
ysis Approach, Amer. Math. Monthly 83 (1976), no. 5,
370–375.
[97] D. Tall and M. Katz, A cognitive analysis of Cauchy’s
conceptions of function, continuity, limit, and infini-
tesimal, with implications for teaching the calculus,
A career at NSA is no ordinary job. It’s a profession dedicated to Educational Studies in Mathematics, to appear.
identifying and defending against threats to our nation. It’s a See http://dx.doi.org/10.1007/s10649-014-
dynamic career filled with challenging and highly rewarding work 9531-9 and http://arxiv.org/abs/1401.1468.
that you can’t do anywhere else but NSA. [98] P. Tannery and C. Henry, Oeuvres de Fermat, Vol. 1,
Gauthier-Villars, 1891.
You, too, can rise above the ordinary. Whether it’s producing [99] , Oeuvres de Fermat, Vol. 3, Gauthier-Villars,
valuable foreign intelligence or preventing foreign adversaries from 1896.
accessing sensitive or classified national security information, you [100] T. Tao, see the post http://terrytao.wordpress.
can help protect the nation by putting your intelligence to work. com/2013/11/16/qualitative-probability-
theory-types-and-the-group-chunk-and-
NSA offers a variety of career fields, paid internships, co-op and group-configuration-theorems/.
scholarship opportunities. [101] T. Tho, Equivocation in the foundations of Leibniz’s
infinitesimal fictions, Society and Politics 6 (2012),
Learn more about NSA and how your career can make a difference
no. 2, 70–98.
for us all.
[102] A. Weil, Number Theory, An Approach through
KNOWINGMATTERS History, From Hammurapi to Legendre, Birkhäuser
Boston, Inc., Boston, MA, 1984.
[103] J. Wheeler, At home in the universe, Masters of Mod-
ern Physics, American Institute of Physics, Woodbury,
NY, 1994.
[104] N. Wiener, Differential space, Journal of Mathemati-
cal Physics 2 (1923), no. 1, 131–174.
Excellent Career [105] T. Williamson, How probable is an infinite sequence
Opportunities for of heads?, Analysis 67 (2007), no. 3, 173–180.
Experts in the Following: [106] W. Wisan, The new science of motion: A study of
Galileo’s De motu locali, Arch. History Exact Sci. 13
n Number Theory (1974), 103–306.
n Probability Theory
n Group Theory
n Finite Field Theory
n Combinatorics
n Linear Algebra
>> Plus other

opportunities WHERE INTELLIGENCE GOES TO WORK®
Search NSA to Download
U.S. citizenship is required. NSA is an Equal Opportunity Employer.

12BWNS-03_3.25x9.5.indd 1 7/21/14 3:31 PM

The True History of The Limit

Uploaded by

Copyright:

Available Formats

The True History of The Limit

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The True History of The Limit

Uploaded by

Copyright:

Available Formats

Fermat, Leibniz, Euler, and the

Gang: The True History of the

848 Notices of the AMS Volume 61, Number 8

technical usage); an infinitesimal is smaller (in

September 2014 Notices of the AMS 849

850 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 851

852 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 853

854 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 855

856 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 857

858 Notices of the AMS Volume 61, Number 8

stated as follows. Denote by x the halo of x, i.e.,

September 2014 Notices of the AMS 859

860 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 861

862 Notices of the AMS Volume 61, Number 8

September 2014 Notices of the AMS 863

Rise Above the Ordinary

>> Plus other

Search NSA to Download

U.S. citizenship is required. NSA is an Equal Opportunity Employer.

864 Notices of the AMS Volume 61, Number 8

You might also like

>> Plus other