Professional Documents
Culture Documents
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The University of Chicago Press and The History of Science Society are collaborating with JSTOR to digitize,
preserve and extend access to Isis.
http://www.jstor.org
Fluxions,
Infinite
Limits,
and
Littlenesse
A Studyof Newton'sPresentation
of the Calculus
By Philip Kitcher*1
I. INTRODUCTION
THE
33
34
PHILIP KITCHER
35
36
PHILIP KITCHER
Newton failed to convince his contemporaries that he had solved the problems of
justification. In the eighteenth century the conflict between geometry and algebra for
the place of basic discipline of mathematics resolved in favor of algebra, and Newton's faithful successors became the proponents of an outworn way of pursuing their
subject.
II. THE NEED FOR GENERALIZATIONAND SIMLIFICATION
37
cases, a natural preliminaryto the extension of analytic techniques. From this point on
(late 1664) his writings showed a clear structurein which the provision of an algorithm
was a fundamental component. The work which is considered central to the development of the calculus was structuredin the following way:
1. An algorithm or method for solving the problem being considered was given.
2. The validity of the algorithm was "demonstrated" by justifying the method for
particular cases. Newton asserted that the method of "proof" revealed how the
algorithm worked for any problem selected from the original problem class.
Newton typically expressed his algorithms in the form of a set of instructions to the
reader. The mechanical application of his rules to the problems at hand would then
yield their solutions. For example, in the early paper on "The General Problem of
Tangents and CurvatureResolved for Algebraic Curves" Newton offered a number of
what he called "Universall theorems." Typical in style is the following method for
finding the tangent to an algebraic curve via the computation of the subnormal v.
Havingye natureof a crookedline expressedin Algebraicalltermeswch are not put one
pte equallto anotherbut all of ym equallto nothing,if each of the termesbe multiplied
by soe many units as x hath dimensionsin them. & then multipliedby y & dividedby x
they shall be a numerator:Also if the signes be changed& each termebe multipliedby
soe many units as y hath dimensionsin yt terme & yn divided by y they shall bee a
denominatorin ye valorof v.14
In order to give an algorithm for the computation of curvaturelater in the same paper,
Newton was forced to introduce expressions of even greater complexity.'5 Consideration of the algorithm quoted above reveals a problem confronting mathematicians
working in the algorithmic tradition: there is no indication of where the rule of thumb
has come from, nor is any connection made between the problem to which this algorithm is directed and other analytic problems. Furthermore, the basis on which the
methods rest consists of several examples in which the result given by the algorithm is
yielded as the solution to the problem when it is set up in terms of infinitesimals.'6 An
example of this type of justification will be given in Section IV. Here it is enough to
note that Newton's resolution of the general problems of tangents and curvature poses
two questions:
1. How are the methods to be reduced to their simplest and most convenient form?
2. How are they to be justified?
We shall first see how Newton's theory of fluxions answers the first question.
IH. THE SIMPLIFYINGPOWER OF FLUXIONS
A search for the way in which Newton's algorithms developed reveals a twofold
progression. In the first place he extended the available techniques to make it possible
to solve more problems from each class, accomplishing this by the device of expanding
"1 See MPN, p. 276. The algorithm is clearly
relatedto the earlierrule of de Sluse.
15Ibid., pp. 289-290. The relative complexity
of the curvaturealgorithmscompared with the
tangentalgorithmis much the same as that of the
38
PHILIP KITCHER
obstinate functions into infinite series.'7 More significant for our study is the way in
which the algorithms are structured in those papers where Newton presented his
method in terms of the fluxion concept. Here he achieved a quite different advance by
reducing the number of classes of problems. Newton replaced the straightforwardbut
cumbersome rules (such as that quoted above) by two fundamental algorithms giving
solutions to two general classes of problems. The solutions to these problems-which
are inversely related-can then be applied to solve any of the old problems. The
method of fluxions enhanced analysis by drawing all the traditional questions together
in a neat and satisfying way.
In Newton's "Method of Fluxions" we find a classic statement of the effectiveness of
the fluxion concept at the algorithmic level.
Now in order to this, I shall observe that all the difficultieshereof (the problems
traditionallyassociatedwiththe analyticart]maybe reducedto thesetwo problemsonly,
which I shall propose,concerninga Space describ'dby local Motion, any how acceleratedor retarded.
I. The length of the space describ'dbeingcontinually(that is, at all times)given;to find
the velocityof the motion at anytimepropos'd.
II. The velocity of the motion being continuallygiven; to find the length of the Space
describ'dat anytimepropos'd.18
Newton went on to offer algorithmic solutions to the first of these problems and, insofar as he could, to the second.19We then see how the problems confronting the method
of analysis are systematically reducible to the fundamental problems. Tangent and
curvatureproblems can be formulated in terms of the first problem, that of finding the
"fluxions." The task of computing quadraturereduces to the second problem.
To appreciate the way in which the reduction takes place we need to understand
Newton's employment of a kinematic conception of curves. The strangeness for us of
Newton's relation of motion to geometry lies in its clash with our view of curves as
given by sets of points in Cartesian space. Ironically, Descartes himself would have
found Newton's viewpoint more sympathetic than ours, since he, like Fermat, constructed geometry on a uniaxial approach.20That is, a curve would be seen as generated
by the motion of an ordinate segment as its foot moved along a base line. This conception loomed large in Isaac Barrow's GeometricalLectures21-especially in the second17 See De Analysi and "Method of Fluxions"
(both in MWN). It is quite clear that Newton's
use of infinite series belongs to analytic mathematics. These series had no meaning in traditional terms until Newton's developmentof the
methodof ratios.
18 "Methodof Fluxions,"MWN, pp. 48-49.
19 The first statement of these problems is in
the work which Whitesideentitles "The Calculus
becomes an Algorithm." Whiteside dates this
tentativelyas having been written in the middle
of 1665. For the statement of the problems see
MPN, p. 344:
1. If two bodys c,d describeye streight lines
ac, bd in ye same time, (callingac = x & bd =
FLUXIONS,LIMITS,INFINITELITTLENESSE
39
and was perhaps one point on which Barrow influenced Newton. Like Barrow, Newton used both the kinematic approach and the static view of a curve as a set of points
in certain geometrical or algebraic relations. Barrow's seeming lack of concern with
comparing the two viewpoints is also similar to the pragmatic approach of Newton's
early work, suggesting that neither man saw any reason except convenience to prefer
one or the other. Later, however, the kinematic emphasis on continuity led Newton to
adopt this conception in his attempt to provide a synthetic foundation for the calculus.
We can now see more easily how Newton reduced the traditional problems to the
fundamental motion problems. The following statement links the tangent problem to
the first motion problem while also making a more general connection:
In ye descriptionof any Mechanicallline what ever, there may bee found two such
motionswch compoundor make up ye motion of ye point describeingit, whose motion
beingby themfoundby ye Lemma,its determinaconshall bee in a tangentto ye mechanicallline.22
The reduction works as follows. Suppose that we wish to find the tangent to the curve
f(x,y) -0 at the point (X, Y).23 Consider the curve as swept out by the motion of a
point, and resolve the motion along the axes. Let the component of velocity along the
x-axis be p(t), that along the y-axis q(t), and let the components when the point is at
(X, Y) be P, Q respectively. Applying the algorithm for the first fundamental problem,
we derive from the general relation f(x,y) 0 a general relation g(x,y,p,q) = 0. In
particular, we have g(X, Y,P,Q) = 0, giving us a relation between P and Q. Let a be the
angle which the tangent at (X, Y) makes with the x-axis. Since the instantaneous
velocity of the moving point at (X, Y) is in the direction of the tangent, we know that
tan a = Q/P. We calculate this ratio from the equation g(X, Y,P,Q) - 0, thus solving
our problem.24
This example is typical of the way in which the kinematic conception of curves
enabled Newton to break down such traditional problems as curvature and quadrature into two stages. In the first stage he showed how the problem could be reduced to
one of the fundamental problems, and the solution to the latter then gave what was
required. The first part of the process represented a natural and direct resolution of
the problem in kinematic terms. Only at the second stage was there need for algorithmic work to be done. A unified domain of study emerged as the connections which
Newton forged with the fundamental problems revealed that the traditional analytic
questions could be seen as naturally related.
The algorithmic reformulation in terms of fluxions thus solved our problem 1 which
we posed for Newton above. Yet it in no way helped with question 2, for nothing had
been done to eliminate the crudeness of the infinitesimal justifications for the algorithms. Indeed, Newton continued to justify his solution of the fundamental problems in terms of infinitesimals. The fluxional approach thus depended on more tradi22 MPN, p. 377. The lemma referredto is the
parallelogramof velocitieslenmna.
23 We considerthe curve as given in Cartesian
coordinates.Problemsfor the method arisewhen
othersystemsareused.
24 So far there is no guaranteethat the ratio
Q:P can be found from the equationg(X, Y,P,Q)
=0. The fact that it can, only emergeswhen we
40
PHILIP KITCHER
of
Letterto Keill, May 15, 1714 (Correspondence
Newtonand Cotes,p. 176): "ffluxions& moments
are quantities of a differentkind. ffluxions are
finitemotions, momentsare infinitelylittle parts.
I put letters with pricks for fluxions, & multiply
fluxions by the letter o to make them become
infinitelylittle...."
27 See MWN, p. xii.
41
42
PHILIP KITCHER
Barrow had no doubt about the viability of our second alternative. Yet, although
the methods of analysis might not be mathematical, Barrow allowed that the problems to which they were addressed had a clear mathematical pedigree. Archimedes had
shown how to find the quadrature of the parabola in a way which was truly geometrical. Barrow attempted to follow directly in the synthetic tradition, solving the
problems of tangents and quadrature by purely geometrical means. His approach led
not to an impressive battery of methods for the working mathematician but to the
perception of relationships among the various classes of problems which his analytically minded predecessors had missed. Barrow's geometrical formulation prevented the
relations from standing out as perspicuously as they do in Newton's work,29but despite this, J. M. Child's eulogy of Barrow is not altogether inappropriate.30For the
combination of the analytic approach with Barrow's drive toward geometrical synthesis and relation of problems created the calculus. This combination was Newton's.3'
The disrespect which Barrow felt for the method of analysis must be opposed, not to
a feeling of confidence on the part of the analysts, but rather to a vague unease. Even
in the work of the foremost practitioners of the analytic art we find genuine consciousness of the ultimate need for a synthetic foundation. The provision of algorithms, the
justification of those algorithms by means of infinitesimals, even the relating of the
problems-all this still left analysis short of the Euclidean ideal. In order to bring the
discipline into line with mathematics, geometrical proof had to be given.32 Newton
took the job of providing such proof seriously.33
The analyst, like the physicist, is interested in methods which are as general and
powerful as possible. Qua analyst or qua physicist, Newton was concerned with
our problem 1-simplification-and he resolved it as we have described. Qua synthesist, however, he wanted to give those methods a firm mathematical underpinning. That
is not to say that he was sensitive to the kind of foundational studies which a twentiethcentury philosopher might favor. It is unclear whether there is any evidence of Newton
29 Thus, although Barrow formulatedwhat is
known as the fundamental theorem of the
calculus, his geometricalversion cannot be seen
as having anything to do with the calculus as
developedby his contemporariesand by Newton.
30 Geometrical Lectures, ed. Child,pp. vii ff.
31 I do not want to suggestanythingmore than
the possibilitythat Newton knew about some of
the very general featuresof Barrow'sapproach.
He may have learned that Barrowregardedthe
problems of the calculus as being related. The
differencein the modes of exhibitionof those relations favored by the two men belies a stronger
connection.In any event, it now seems clear that
Barrow'sinfluenceon Newton was nowherenear
as greatas was once believed.
32 In equating the synthetic method of proof
with reduction to geometry I am following a
conflationwhichNewton seemsto make.Newton
may have been impervious to the distinction
between the form of a proof and the content of
that proof, a distinction which we make with
ease, and it may thus be that his foundational
43
asking himself what o denotes (i.e., what an infinitesimal is). Indeed, in the light of his
De Quadraturawith its instrumentalist attitude toward infinitesimals, the question
would seem to be meaningless for him. In pursuing his own brand of foundational
work Newton shows instead the respect for geometry which had traditionally been
common but which in his lifetime was beginning to be questioned.
The extension of the analytic art was necessarily Newton's first concern. Once the
method had been elegantly formulated, Newton turned to the problem of providing
rigorous proof. The need for intermediatejustification emerges from an understanding
of the temper of the times: the Descartes/Fermat controversy is symptomatic of the
fact that jealous rivals had to be convinced. Newton was not required to invent this
temporary substitute for himself, however, for there was already a flourishing tradition
of justifying results using infinitesimals. By such means certain algebraic manipulations could be renderedmore comprehensible and perhaps a little less implausible. The
status of such justification is that of a warranty offered by a company we do not quite
trust-it provides an extra safeguard against trouble, but it fails to give complete
assurance.
At an early stage of his career Newton gave several hints that he was not satisfied to
let justification rest with the method of infinitesimals. It was obviously natural for him
to ignore these worries until analysis had reached its goal. Unlike Barrow, Newton was
able to see what could and should be done with analysis; for him the synthetic pull was
not as urgent. That pull did not drag him into the puristic complexities of Barrow's
work, but, as we shall see, neither did it leave him unmoved, even while analysis and
the development of algorithms were his main concerns.
To show the way in which the method of infinitesimaljustification falls short of the
geometric paradigm of proof, and thus how our question 2 (How are the methods to
be justified?) arises, we must examine a case in which Newton used this method. The
following example, taken from the "Method of Fluxions," was intended to vindicate
Newton's algorithm for finding the relation between velocities from the relation between distances by deriving the result yielded by the algorithm in a particularcase.34
Now sincethe moments,as xo and yo are the indefinitelylittle accessionsof the flowing
quantitiesx andy, by whichthose quantitiesareincreasedthroughthe severalindefinitely
smallintervalsof time; it follows that those quantitiesx and y afterany indefinitelysmall
interval of time become x +ko and y +
yo, as between x and y; so that x +Jo and
y + 'o maybe substitutedin the sameequationfor those quantities,insteadof x andy.
Thereforelet any equationx3 -ax2 + axy - y5 zr0 be given,35and substitutex +*o
for x andy +y'o for y, and therewill arise
X3+ 3Rox2+ 3R2oox +303
- ax2 - 2a*ox - a*2oo
+ axy + axoy + ayox + ayxoo
-y3
- 3yoy2
- 320ooy
-303
?O
PHILIP KITCHER
44
2axk -a*2o
FLUXIONS,LIMITS,INFINITELITITLENESSE
45
blind to defects in the method which he used. In the remainder of this section we shall
see how his toleration of the method of infinitesimals could coexist with a realization
of its limitations.
The opening remarks of the "Method of Fluxions" make it clear that Newton was
aware of the distinction between what was permissible in the context of discovery and
what was requiredfor proof. He also showed that he regarded his fluxional calculus to
belong to the former context, to analytic rather than to synthetic mathematics. There
is an implicit value judgment in the way in which he drew the distinction:
Having observ'd that most of our modern Geometriciansneglecting the synthetical
method of the Ancients,havingappliedthemselveschieflyto the analyticalArt, and by
the help of it haveovercomeso manyand so greatDifficulties,that all the Speculationsof
Geometryseemto be exhausted,exceptthe Quadratureof Curves,and some otherthings
of a like Nature which are not yet broughtto Perfection;To this end I thought it not
amiss,for the sake of young studentsin this Science,to drawup the followingTreatise;
wherinI have endeavouredto enlargethe Boundariesof Analyticks,and to make some
Improvementsin the Doctrineof CurveLines.42
Like Barrow, Newton recognized the ancestry of the problems of the calculus and
noted also the distinction between two ways of approaching them. Here he distinguished his aim: he proposed to present an extension of the new analytic methods.43
With his acknowledgement of a difference between his own enterprise and that of
"the Ancients" we begin to see that Newton could work in the analytic tradition,
using its methods and obeying its code, while allowing the question of how those
methods were to be justified to await completion of his analytic work. His careful
description of his chosen task certainly suggests this.
To find a clear connection between Newton's recognition of the pedigree of the
calculus and the method of infinitesimals we must go back to his work on the resolution of the general problems of tangents and curvature for algebraic curves. In a passage featuring the use of infinitesimals Newton commented, "wch operacon cannot in
this case bee understood to be good unless infinite littlenesses may bee considered
geometrically."44This illustrates the synthetic pull on Newton's thought. Recognizing
the shortcomings of his algebra, he saw the need for a clear interpretation in geometric
terms and saw that this was only partially accomplished by the invocation of the infinitesimal. The method relied upon this interpretation-the analytic methods were
dependent on the geometry-and Newton concluded, even in 1665, that if his justifications were to have the status of full proofs, "infinite littlenesse" must be construed in
geometrical terms.
For further evidence of his view of "Analyticks" as growing out of geometry, we
may turn to passages in Newton's early work where he introduced the concept of the
indefinite integral and where he hinted darkly at the difficulty of integrating differential equations.
That a line may be squaredGeometricallytis requiredyt its area may be expressedin
generallby someequationin wchthereis an unknownequantity....45
p. 36.
He achieved this, as we have remarked,in
two ways. First, with the method of series he
opened up new problemsin each problemclass.
Secondly, with the method of fluxions he pre42 MWN,
43
p. 344.
46
PHILIP KITCHER
If an equation expressingye relation of their motions bee given, tis more difficult&
sometimesGeometricallyimpossible,therebyto findye relationof ye spacesdescribedby
thesemotions.46
The only way to account for the seemingly irrelevant occurrence of references to
geometry in these passages is to construe them as further indications that Newton
considered analysis to be parasitic upon geometry for its problems and ultimately to
be justified in geometrical terms.
Newton was further conscious of using the method of infinitesimals within a definite
tradition. The De Analysi provides an example of this: "Neither am I afraid to speak of
Unity in points, or Lines infinitely small, since Geometers are wont now to consider
Proportions even in such a case, when they make use of the Methods of Indivisibles."47
The first clause of the sentence indicates that Newton was aware of criticisms of and
perhaps defects in the method of infinitesimals. He did not undercut objections by
careful statement and counterargument; instead he dismissed them on pragmatic
grounds, since there was a recent tradition-the fashion of the geometers of the daywhich used such methods, and Newton regarded his work as falling therein. In the De
Analysi, as in the "Method of Fluxions," since the work was intended for circulation,
Newton sagaciously guarded himself by defining his aims in the sphere of the analytic
art. His doubt about the geometric construal of "infinite littlenesse" was confided to
his notebook. Indeeed, Newton's synthetic worries only became public once he had
appeased them.
As we have seen, Newton was quite conscious of the distinction between "Analyticks" and the old synthetic mathematics. Carefully distinguishing his own work, he
remained aware that the method of infinitesimals was only appropriate as a temporary
backing for algebraic maneuvers in the context of discovery. In writing about "infinite
littlenesse" he pointed the way in which to answer criticism and correct defects. Once
the analytic program was accomplished he was to bring the new results within the old
paradigm. Newton's synthetic conscience went on view in the Principia. It seems that it
was developing twenty years earlier.
V. FIRST AND LAST RATIOS AND RIGOROUS JUSTIFICATION
46 Ibid.,p. 302.
47
51Principia, p. 32.
52MWN, p. 141.
53Ibid.
54 Ibid., p. 142.
48
PHILIP KITCHER
He was quite aware of what he had achieved. Immediately after his proof of de
Sluse's rule he pointed out that his method "is agreeable to the Geometry of the Ancients."56In the same way, in the second Scholium to Book I of the Principia he emphasized the geometrical essence of the calculus.57His purpose in the Principia was
however slightly different from that in the later Treatise. In the earlier work he used
the method of first and last ratios to give a geometrical vindication of the method of
infinitesimals. Newton showed in a series of lemmas that all justifications using infinitesimals could be replaced by exact geometrical proofs in terms of ultimate ratios.
The method of ratios acts here as a meta-level principle which supports the fluxional
calculus by demonstrating that the traditional analytic means of vindication is a shorthand for a true mathematical proof. The same theme appears in the Treatise, where we
have the direct demonstration as well.
Two quotations from Newton make clear the motives for his vindication of analysis.
First,
These Lemmasare premisedto avoid the tediousnessof deducinginvolveddemonstraaccordingto the mannerof the ancientgeometers.For demonstrations
tions adabsurdum,
are shorter by the method of indivisibles;but because the hypothesis of indivisibles
seemssomewhatharsh,and thereforethat methodis reckonedless geometrical,I chose to
reducethe demonstrationsof the following Propositionsto the first and last sums and
ratiosof nascentand evanescentquantities,thatis, to the limitsof those sumsand ratios,
and so to premise,as short as I could, the demonstrationsof those limits. For herebythe
samethingis performedas by the methodof indivisibles;and now those principlesbeing
demonstrated,we may use them with greatersafety. Thereforeif hereafterI should
happento considerquantitiesas madeup of particles,or shoulduse little curvedlines for
right ones, I would not be understoodto mean indivisibles,but evanescentdivisible
quantities;not the sums and ratios of determinateparts, but alwaysthe limits of those
sums and ratios; and that the force of such demonstrationsalways depends on the
methodlaid downin the foregoingLemmas.58
The lemmas of Book I-especially Lemma VII-do indeed vindicate the analytic
approach. They show how the method of infinitesimals may be used to justify algorithms so long as it is recognized as a substitute for geometrical proof. Infinitesimals
flout the important considerations of continuity; yet, because they are sanctioned by
the method of limits, they may be used instead of the laborious geometrical proofs. A
parallel with modern mathematics presents itself: in doing formal logic or arithmetic
we extend our vocabulary by means of definitions and our methods of inference by
derived rules. These procedures are theoretically eliminable, but they are legitimized
by our system and are used for simplification.59Infinitesimals ultimately find an analogous function in Newton's theory of the calculus.
5sMWN, p. 143.
57Principia,pp. 37-39. Note esp. p. 39: "And
since such limits are certain and definite, to
determine the same is a problem strictly geometrical." The equation of certainty with geometry gives an interestingfurtherelaborationof
Newton's attitude toward the geometrical
paradigmof mathematics.
58 Ibid.,p. 38.
69See, e.g., Alonzo Church, Introductionto
MathematicalLogic (Princeton:PrincetonUniv-
49