Editor’s note
The first published account of the differential calculus. The Latin follows Dutens, checked against the sense; Leibniz's notation ("aequ." for =, the colon for division, the comma as a bracket, "√" with a vinculum) is regularized into modern form in both columns, and the ambiguous signs ±, (±), ((±)) are kept. Where Dutens's formulas are illegible in the scan, they are reconstructed from the surrounding text and marked. The figure (Tab. VI, figs. 1–3) is not reproduced.
Original
English
1
Nova methodus pro maximis et minimis, itemque tangentibus, quae nec fractas nec irrationales quantitates moratur, et singulare pro illis calculi genus.
A new method for maxima and minima, and also for tangents, which is not hindered by fractional or irrational quantities, and a singular kind of calculus for them.
2
Sit axis AX, et curvae plures, ut VV, WW, YY, ZZ, quarum ordinatae ad axem normales VX, WX, YX, ZX, quae vocentur respective v, w, y, z; et ipsa AX abscissa ab axe vocetur x. Tangentes sint VB, WC, YD, ZE, axi occurrentes respective in punctis B, C, D, E. Jam recta aliqua pro arbitrio assumta vocetur dx, et recta quae sit ad dx ut v (vel w, vel y, vel z) est ad VB (vel WC, vel YD, vel ZE) vocetur dv (vel dw, vel dy, vel dz), sive differentia ipsarum v (vel ipsarum w, aut y, aut z). His positis calculi regulae erunt tales.
Let there be an axis AX and several curves, such as VV, WW, YY, ZZ, whose ordinates normal to the axis, VX, WX, YX, ZX, are called respectively v, w, y, z; and let AX, the abscissa cut off from the axis, be called x. Let the tangents be VB, WC, YD, ZE, meeting the axis respectively at the points B, C, D, E. Now let some straight line taken at will be called dx, and let the line that is to dx as v (or w, or y, or z) is to VB (or WC, or YD, or ZE) be called dv (or dw, or dy, or dz), or the difference of the v's (or of the w's, or y's, or z's). With these assumptions the rules of the calculus will be as follows.
3
Sit a quantitas data constans, erit da aequalis 0, et d(ax) erit aequ. a dx. Si sit y aequ. v (seu ordinata quaevis curvae YY aequalis cuivis ordinatae respondenti curvae VV), erit dy aequ. dv. Jam Additio et Subtractio: si sit z − y + w + x aequ. v, erit d(z − y + w + x), seu dv, aequ. dz − dy + dw + dx. Multiplicatio: d(xv) aequ. x dv + v dx, seu posito y aequ. xv, fiet dy aequ. x dv + v dx. In arbitrio enim est vel formulam, ut xv, vel compendio pro ea literam, ut y, adhibere. Notandum et x et dx eodem modo in hoc calculo tractari, ut y et dy, vel aliam literam indeterminatam cum sua differentiali. Notandum etiam non dari semper regressum a differentiali aequatione, nisi cum quadam cautione, de quo alibi. Porro Divisio: d(v/y), vel (posito z aequ. v/y) dz, aequ. (±v dy ∓ y dv)/yy.
Let a be a given constant quantity; then da will equal 0, and d(ax) will equal a dx. If y equals v (that is, if any ordinate of the curve YY equals any corresponding ordinate of the curve VV), dy will equal dv. Now Addition and Subtraction: if z − y + w + x equals v, then d(z − y + w + x), or dv, will equal dz − dy + dw + dx. Multiplication: d(xv) equals x dv + v dx; or, setting y equal to xv, dy will equal x dv + v dx. For it is at our discretion to use either the formula, such as xv, or for brevity a letter in its place, such as y. Note that x and dx are treated in this calculus in the same way as y and dy, or any other indeterminate letter with its differential. Note also that one cannot always return from a differential equation except with some caution, of which elsewhere. Further, Division: d(v/y), or (setting z equal to v/y) dz, equals (±v dy ∓ y dv)/yy.
4
Quoad Signa hoc probe notandum, cum in calculo pro litera substituitur simpliciter ejus differentialis, servari quidem eadem signa, et pro +z scribi +dz, pro −z scribi −dz, ut ex additione et subtractione paulo ante posita apparet; sed quando ad exegesin valorum venitur, seu cum consideratur ipsius z relatio ad x, tunc apparere an valor ipsius dz sit quantitas affirmativa, an nihilo minor seu negativa: quod posterius cum fit, tunc tangens ZE ducitur a puncto Z non versus A, sed in partes contrarias seu infra X, id est tunc cum ipsae ordinatae z decrescunt crescentibus x. Et quia ipsae ordinatae v modo crescunt, modo decrescunt, erit dv modo affirmativa, modo negativa quantitas, et priore casu ₁V₁B tangens ducitur versus A; posteriore ₂V₂B in partes aversas: neutrum autem fit in medio circa M, quo momento ipsae v neque crescunt neque decrescunt, sed in statu sunt, adeoque fit dv aequ. 0, ubi nihil refert quantitas sitne affirmativa an negativa, nam +0 aequ. −0: eoque in loco ipsa v, nempe ordinata LM, est maxima (vel, si convexitatem axi obverteret, minima), et tangens curvae in M neque supra X ducitur ad partes A, ibique axi propinquat, neque infra X ad partes contrarias, sed est axi parallela. Si dv sit infinita respectu ipsius dx, tunc tangens est ad axem recta, seu est ipsa ordinata. Si dv et dx aequales, tangens facit angulum semirectum ad axem. Si crescentibus ordinatis v crescunt etiam ipsa earum incrementa vel differentiae dv (seu si, positis dv affirmativis, etiam ddv, differentiae differentiarum, sunt affirmativae, vel negativis negativae), curva axi obvertit concavitatem, alias convexitatem: ubi vero est maximum vel minimum incrementum, vel ubi incrementa ex decrescentibus fiunt crescentia aut contra, ibi est punctum flexus contrarii, et concavitas atque convexitas inter se permutantur, modo non et ordinatae ibi ex crescentibus fiant decrescentes vel contra, tunc enim concavitas aut convexitas maneret: ut autem incrementa continuent crescere aut decrescere, ordinatae vero ex crescentibus fiant decrescentes vel contra, fieri non potest. Itaque punctum flexus contrarii locum habet, quando neque v neque dv existente 0, tamen ddv est 0. Unde etiam problema flexus contrarii non duas, ut problema maximae, sed tres habet radices aequales. Atque haec omnia quidem pendent a recto usu signorum.
As for the signs, this must be carefully noted: when in the calculus a letter is simply replaced by its differential, the same signs are indeed kept, and for +z we write +dz, for −z we write −dz, as appears from the addition and subtraction set out just above. But when we come to the evaluation of the values, that is, when the relation of z to x is considered, it then appears whether the value of dz is an affirmative quantity or less than nothing, that is, negative. When the latter happens, the tangent ZE is drawn from the point Z not towards A but in the opposite direction, below X; that is, when the ordinates z decrease as the x's increase. And because the ordinates v sometimes increase and sometimes decrease, dv will be sometimes an affirmative and sometimes a negative quantity; in the first case the tangent ₁V₁B is drawn towards A, in the second ₂V₂B in the opposite direction. But neither happens in the middle, around M, at the moment when the v's neither increase nor decrease but are at a standstill, so that dv becomes equal to 0, where it makes no difference whether the quantity is affirmative or negative, for +0 equals −0. At that place v, namely the ordinate LM, is a maximum (or, if the curve turned its convexity to the axis, a minimum), and the tangent of the curve at M is drawn neither above X towards A, approaching the axis there, nor below X in the opposite direction, but is parallel to the axis. If dv is infinite with respect to dx, the tangent is perpendicular to the axis, that is, it is the ordinate itself. If dv and dx are equal, the tangent makes half a right angle with the axis. If, as the ordinates v increase, their increments or differences dv also increase (that is, if, the dv's being affirmative, the ddv's, the differences of the differences, are also affirmative, or negative when they are negative), the curve turns its concavity towards the axis; otherwise its convexity. But where the increment is a maximum or minimum, or where the increments change from decreasing to increasing or the reverse, there is a point of contrary flexure, and concavity and convexity are exchanged, provided the ordinates do not also change there from increasing to decreasing or the reverse, for then the concavity or convexity would remain. But it is impossible for the increments to continue to increase or decrease while the ordinates change from increasing to decreasing or the reverse. Therefore a point of contrary flexure occurs when, neither v nor dv being 0, ddv is nevertheless 0. Hence the problem of contrary flexure has not two equal roots, as the problem of the maximum has, but three. And all this depends on the correct use of the signs.
5
Interdum autem adhibenda sunt Signa ambigua, ut nuper in divisione, antequam scilicet constet quomodo explicari debeant. Et quidem si crescentibus x crescunt (decrescunt) v/y, debent signa ambigua in d(v/y), seu in (±v dy ∓ y dv)/yy, ita explicari, ut haec fractio fiat quantitas affirmativa (negativa). Significat autem ∓ contrarium ipsius ±, ut si hoc sit +, illud sit −, vel contra. Possunt et in eodem calculo occurrere plures ambiguitates, quas distinguo parenthesibus, exempli causa, si esset v/y + y/z + x/v aequ. w, foret (±v dy ∓ y dv)/yy + ((±) y dz (∓) z dy)/zz + (((±)) x dv ((∓)) v dx)/vv aequ. dw: alioqui ambiguitates ex diversis capitibus ortae confunderentur. Ubi notandum signum ambiguum in se ipsum ductum dare +, in suum contrarium dare −, in aliud ambiguum formare novam ambiguitatem ex ambabus dependentem.
Sometimes, however, ambiguous signs must be used, as just now in division, before it is established how they are to be resolved. And indeed, if v/y increases (decreases) as the x's increase, the ambiguous signs in d(v/y), that is, in (±v dy ∓ y dv)/yy, must be resolved so that this fraction becomes an affirmative (negative) quantity. And ∓ signifies the contrary of ±, so that if the one is +, the other is −, or the reverse. Several ambiguities may also occur in the same calculation, which I distinguish by parentheses; for example, if v/y + y/z + x/v were equal to w, we would have (±v dy ∓ y dv)/yy + ((±) y dz (∓) z dy)/zz + (((±)) x dv ((∓)) v dx)/vv equal to dw; otherwise ambiguities arising from different sources would be confused. Note here that an ambiguous sign multiplied by itself gives +, multiplied by its contrary gives −, and multiplied by another ambiguous sign forms a new ambiguity depending on both.
6
Potentia: d(xᵃ) = a xᵃ⁻¹ dx, exempli gratia d(x³) = 3x² dx. d(1/xᵃ) = −a dx/xᵃ⁺¹, exempli gratia si w sit 1/x³, fiet dw = −3dx/x⁴. Radices: d(ᵇ√xᵃ) = (a/b) dx ᵇ√(xᵃ⁻ᵇ). Hinc d√y = dy/(2√y); nam eo casu a est 1 et b est 2; ergo (a/b) ᵇ√(xᵃ⁻ᵇ) est ½ √(y⁻¹); jam y⁻¹ idem est quod 1/y, ex natura exponentium progressionis Geometricae, et √(1/y) est 1/√y. d(1/ᵇ√xᵃ) = −a dx/(b ᵇ√(xᵃ⁺ᵇ)). Suffecisset autem regula potentiae integrae tam ad fractas quam ad radices determinandas; potentia enim fit fracta cum exponens est negativus, et mutatur in radicem cum exponens est fractus: sed malui consequentias istas ipse deducere quam aliis deducendas relinquere, cum sint admodum generales et crebro occurrentes, et in re per se implicita praestet facilitati consulere.
Power: d(xᵃ) = a xᵃ⁻¹ dx; for example, d(x³) = 3x² dx. d(1/xᵃ) = −a dx/xᵃ⁺¹; for example, if w is 1/x³, then dw = −3dx/x⁴. Roots: d(ᵇ√xᵃ) = (a/b) dx ᵇ√(xᵃ⁻ᵇ). Hence d√y = dy/(2√y); for in this case a is 1 and b is 2, so (a/b) ᵇ√(xᵃ⁻ᵇ) is ½ √(y⁻¹); now y⁻¹ is the same as 1/y, from the nature of the exponents of a geometric progression, and √(1/y) is 1/√y. d(1/ᵇ√xᵃ) = −a dx/(b ᵇ√(xᵃ⁺ᵇ)). The rule for integral powers would have sufficed for determining both fractions and roots, for a power becomes a fraction when the exponent is negative and changes into a root when the exponent is fractional; but I preferred to draw these consequences myself rather than leave them for others to draw, since they are very general and occur often, and in a matter intricate in itself it is better to consult ease.
7
Ex cognito hoc velut Algorithmo, ut ita dicam, calculi hujus, quem voco differentialem, omnes aliae aequationes differentiales inveniri possunt per calculum communem, maximaeque et minimae, itemque tangentes haberi, ita ut opus non sit tolli fractas aut irrationales aut alia vincula, quod tamen faciendum fuit secundum Methodos hactenus editas. Demonstratio omnium facilis erit in his rebus versato, et hoc unum hactenus non satis expensum consideranti, ipsas dx, dy, dv, dw, dz ut ipsarum x, y, v, w, z (cujusque in sua serie) differentiis sive incrementis vel decrementis momentaneis proportionales haberi posse. Unde fit ut, proposita quacunque aequatione, scribi possit ejus aequatio differentialis; quod fit pro quolibet membro (id est parte quae sola additione vel subtractione ad aequationem constituendam concurrit) substituendo simpliciter quantitatem membri differentialem; pro alia vero quantitate (quae non ipsa est membrum, sed ad membrum formandum concurrit) ejus quantitatem differentialem ad formandam quantitatem differentialem ipsius membri adhibendo, non quidem simpliciter, sed secundum Algorithmum hactenus praescriptum. Editae vero hactenus Methodi talem transitum non habent; adhibent enim plerumque rectam ut DX vel aliam hujusmodi, non vero rectam dy, quae ipsis DX, XY, dx est quarta proportionalis, quod omnia turbat; hinc praecipiunt ut fractae et irrationales (quas indeterminatae ingrediuntur) prius tollantur. Patet etiam methodum nostram porrigi ad lineas transcendentes, quae ad calculum Algebraicum revocari non possunt, seu quae nullius sunt certi gradus, idque universalissimo modo, sine ullis suppositionibus particularibus non semper succedentibus, modo teneatur in genere tangentem invenire esse rectam ducere quae duo curvae puncta distantiam infinite parvam habentia jungat, seu latus productum polygoni infinitanguli, quod nobis curvae aequivalet. Distantia autem illa infinite parva semper per aliquam differentialem notam, ut dv, vel per relationem ad ipsam exprimi potest, hoc est per notam quandam tangentem. Speciatim, si esset y quantitas transcendens, exempli causa ordinata cycloeidis, eaque calculum ingrederetur cujus ope ipsa z, ordinata alterius curvae, esset determinata, quaereretur dz, seu per eam tangens hujus curvae posterioris, utique determinanda esset dz per dy; haberetur autem dy, quia habetur tangens cycloeidis. Ipsa autem tangens cycloeidis, si nondum haberi fingeretur, similiter calculo inveniri posset ex data proprietate tangentium circuli.
From this, as it were, Algorithm, so to speak, of this calculus, which I call differential, once it is known, all other differential equations can be found by the common calculus, and maxima and minima, and likewise tangents, can be obtained, without any need to remove fractions, irrationals or other bonds, as had to be done according to the methods published hitherto. The demonstration of all this will be easy for one versed in these matters who considers this one point, not sufficiently weighed until now: that dx, dy, dv, dw, dz can be taken as proportional to the momentary differences, that is, increments or decrements, of x, y, v, w, z (each in its own series). Hence, given any equation, its differential equation can be written. This is done by simply substituting, for each term (that is, each part that concurs in constituting the equation by addition or subtraction alone), the differential quantity of the term; but for any other quantity (which is not itself a term but concurs in forming a term), by using its differential quantity to form the differential quantity of the term itself, not simply, but according to the Algorithm prescribed above. The methods published hitherto have no such transition; for they mostly use a line such as DX or another of that kind, and not the line dy, which is the fourth proportional to DX, XY and dx, and this throws everything into confusion; hence they prescribe that fractions and irrationals (into which the indeterminates enter) be removed first. It is clear also that our method extends to transcendental lines, which cannot be reduced to algebraic calculation, that is, which are of no definite degree, and this in the most universal way, without any particular assumptions that do not always succeed; provided it is held in general that to find a tangent is to draw a straight line joining two points of the curve at an infinitely small distance, that is, the extended side of the infinitangular polygon which for us is equivalent to the curve. That infinitely small distance can always be expressed by some known differential, such as dv, or by a relation to it, that is, by some known tangent. Specifically, if y were a transcendental quantity, for example the ordinate of a cycloid, and it entered a calculation by which z, the ordinate of another curve, was determined, and dz were sought, or through it the tangent of this second curve, then dz would certainly have to be determined through dy; and dy would be had, because the tangent of the cycloid is had. And the tangent of the cycloid itself, if we supposed it not yet known, could likewise be found by the calculus from the known property of the tangents of the circle.
Original
English
8
Placet autem exemplum calculi proponere, ubi notetur me divisionem hic designare hoc modo, x:y, quod idem est ac x divis. per y, seu x/y. Sit aequatio prima seu data: x/y + (a + bx)(c − xx)/(ex + fxx)² + ax√(gg + yy) + yy/√(hh + lx + mxx) aequ. 0, exprimens relationem inter x et y, seu inter AX et XY, positis ipsis a, b, c, e, f, g, h, l, m esse datas; quaeritur modus ex dato puncto Y educendi YD quae curvam tangat, seu quaeritur ratio rectae DX ad rectam datam XY. Compendii causa pro a + bx scribamus n; pro c − xx, p; pro ex + fxx, q; pro gg + yy, r; pro hh + lx + mxx, s. Fiet x/y + np/qq + ax√r + yy/√s aequ. 0, quae sit Aequatio Secunda. Jam ex calculo nostro constat d(x/y) esse (±x dy ∓ y dx)/yy; similiter d(np/qq) esse [(±)2np dq (∓) q(n dp + p dn)]/q³; et d(ax√r) esse ax dr/(2√r) + a dx √r; et d(yy/√s) esse [((±)) yy ds ((∓)) 4sy dy]/(2s√s): quae omnes quantitates differentiales, inde ab ipso d(x/y) usque ad d(yy/√s), in unum additae facient 0, et dabunt hoc modo Aequationem tertiam; ita enim pro membris secundae aequationis substituuntur quantitates eorum differentiales. Jam dn est b dx, dp est −2x dx, dq est e dx + 2fx dx, dr est 2y dy, et ds est l dx + 2mx dx. Quibus valoribus in Aequatione tertia substitutis habebitur Aequatio quarta, ubi quantitates differentiales, quae solae supersunt, nempe dx, dy, semper reperiuntur extra nominatores et vincula, et unumquodque membrum afficitur vel per dx vel per dy, servata semper lege homogeneorum quoad has duas quantitates, quomodocunque implicatus sit calculus: unde semper haberi potest valor ipsius dx/dy, seu rationis dx ad dy, hoc est DX quaesitae ad XY datam; quae ratio in hoc nostro calculo (mutando aequationem quartam in Analogiam) erit ut ∓x/yy − axy/√r ((±)) 2y/√s est ad ±1/y (±) 2np(e + 2fx)/q³ (∓) (−2nx + pb)/qq + a√r ((±)) yy(l + 2mx)/(2s√s). Dantur autem x et y ex dato puncto Y. Dantur et valores supra scripti literarum n, p, q, r, s per x et y. Habetur ergo quaesitum. Atque hoc exemplum satis implicatum ideo tantum adscripsimus, ut modus superioribus regulis in calculo etiam difficiliore utendi appareret. Nunc praestat usum in exemplis intellectui magis obviis ostendere.
I should like to give an example of the calculus, noting that I designate division here thus, x:y, which is the same as x divided by y, or x/y. Let the first or given equation be x/y + (a + bx)(c − xx)/(ex + fxx)² + ax√(gg + yy) + yy/√(hh + lx + mxx) = 0, expressing the relation between x and y, that is, between AX and XY, where a, b, c, e, f, g, h, l, m are given. The method is sought for drawing from the given point Y the line YD that touches the curve; that is, the ratio of the line DX to the given line XY is sought. For brevity, let us write n for a + bx; p for c − xx; q for ex + fxx; r for gg + yy; s for hh + lx + mxx. We get x/y + np/qq + ax√r + yy/√s = 0, which is the second equation. Now from our calculus it is established that d(x/y) is (±x dy ∓ y dx)/yy; likewise that d(np/qq) is [(±)2np dq (∓) q(n dp + p dn)]/q³; that d(ax√r) is ax dr/(2√r) + a dx √r; and that d(yy/√s) is [((±)) yy ds ((∓)) 4sy dy]/(2s√s). All these differential quantities, from d(x/y) to d(yy/√s), added together make 0, and so give the third equation; for in this way the differential quantities of the terms of the second equation are substituted for those terms. Now dn is b dx, dp is −2x dx, dq is e dx + 2fx dx, dr is 2y dy, and ds is l dx + 2mx dx. Substituting these values in the third equation, we obtain a fourth, in which the differential quantities, which alone remain, namely dx and dy, are always found outside denominators and bonds, and every term is affected either by dx or by dy, the law of homogeneity with respect to these two quantities being always preserved, however complicated the calculation. From this the value of dx/dy can always be had, that is, the ratio of dx to dy, that is, of the sought DX to the given XY. In our calculation this ratio (turning the fourth equation into a proportion) will be as ∓x/yy − axy/√r ((±)) 2y/√s is to ±1/y (±) 2np(e + 2fx)/q³ (∓) (−2nx + pb)/qq + a√r ((±)) yy(l + 2mx)/(2s√s). [The signs of this proportion are partly illegible in Dutens and are given as they follow from the fourth equation.] Now x and y are given from the given point Y; and the values of the letters n, p, q, r, s written above are given through x and y. So the thing sought is had. We have added this rather complicated example only to show the way of using the above rules even in a more difficult calculation. It is now better to show their use in examples more obvious to the understanding.
9
Data sint duo puncta C et E, et recta SS in eodem cum ipsis plano; quaeritur punctum F in recta SS ita sumendum, ut junctis CF, FE, sit aggregatum rectangulorum CF in datam h et FE in datam r omnium possibilium minimum; hoc est, si SS sit mediorum separatrix, et h repraesentet densitatem medii, ut aquae, a parte C, et r densitatem medii, ut aeris, a parte E, quaeritur punctum F tale ut via a C ad E per F sit omnium possibilium facillima. Ponamus omnia illa rectangulorum aggregata possibilia, vel omnes viarum possibilium difficultates, repraesentari per ipsas KV, curvae VV ordinatas ad rectam GK normales, quas vocabimus ω, quaerique minimam earum, NM. Quia dantur puncta C et E, dabuntur et perpendiculares ad SS, nempe CP (quam vocabimus c) et QE (quam e), et praeterea PQ (quam p); ipsam autem QF, quae sit aequalis ipsi GN (vel AX), vocabimus x, et CF, f, et EF, g; fiet FP, p − x; f aequ. √(cc + pp − 2px + xx), seu compendio √l; et g aequ. √(ee + xx), seu compendio √m. Habemus ergo ω aequ. h√l + r√m, cujus aequationis aequatio differentialis (posito dω esse 0, in casu minimae) est 0 aequ. h dl/(2√l) + r dm/(2√m), per regulas calculi nostri traditas; jam dl est −2dx(p − x), et dm est 2x dx; ergo fit h(p − x)/f aequ. rx/g. Quod si jam haec accommodentur ad dioptricam, et ponantur f et g, seu CF et EF, aequales, quia eadem manet refractio in puncto F quantacunque ponatur longitudo rectae CF, fiet h(p − x) aequ. rx, seu h : r :: x : p − x, seu h ad r ut QF ad FP; hoc est, sinus angulorum incidentiae et refractionis, FP et QF, erunt reciproce ut r et h, densitates mediorum in quibus fit incidentia et refractio. Quae densitas tamen non respectu nostri, sed respectu resistentiae quam radiis lucis faciunt, intelligenda est. Et habetur ita demonstratio calculi alibi a nobis in his ipsis Actis exhibiti, quando generale Opticae, Catoptricae et Dioptricae fundamentum exponebamus. Cum alii doctissimi Viri multis ambagibus venati sint, quae hujus calculi peritus tribus lineis imposterum praestabit.
Let two points C and E be given, and a straight line SS in the same plane with them. A point F is sought on the line SS, to be taken so that, when CF and FE are joined, the sum of the rectangles of CF by the given h and FE by the given r is the least of all possible. That is, if SS is the boundary between two media, and h represents the density of the medium on the side of C, such as water, and r the density of the medium on the side of E, such as air, a point F is sought such that the path from C to E through F is the easiest of all possible. Let us suppose that all these possible sums of rectangles, or all the difficulties of the possible paths, are represented by the lines KV, ordinates of the curve VV normal to the line GK, which we shall call ω, and that the least of them, NM, is sought. Since the points C and E are given, the perpendiculars to SS are also given, namely CP (which we shall call c) and QE (which we shall call e), and besides PQ (which we shall call p); QF, which is equal to GN (or AX), we shall call x; CF, f; and EF, g. Then FP is p − x; f equals √(cc + pp − 2px + xx), or for short √l; and g equals √(ee + xx), or for short √m. We have therefore ω = h√l + r√m, the differential equation of which (setting dω = 0, in the case of the minimum) is 0 = h dl/(2√l) + r dm/(2√m), by the rules of our calculus given above. Now dl is −2dx(p − x) and dm is 2x dx; so we get h(p − x)/f = rx/g. Now if this is applied to dioptrics, and f and g, that is, CF and EF, are taken equal, since the refraction at the point F remains the same however great the length of CF is taken, we get h(p − x) = rx, or h : r :: x : p − x, that is, h is to r as QF to FP; that is, the sines of the angles of incidence and refraction, FP and QF, will be reciprocally as r and h, the densities of the media in which the incidence and refraction take place. This density, however, is to be understood not with respect to us but with respect to the resistance the media offer to the rays of light. And so we have the demonstration of the calculation that we gave elsewhere in these same Acta when we set out the general foundation of optics, catoptrics and dioptrics. Other very learned men have hunted for this by many roundabout ways; one skilled in this calculus will henceforth achieve it in three lines.
10
Quod alio adhuc exemplo docebo. Sit curva 133 talis naturae, ut a puncto ejus quocunque, ut 3, ductae ad sex puncta fixa in axe posita, 4, 5, 6, 7, 8, 9, sex rectae 34, 35, 36, 37, 38, 39 simul additae sint rectae datae g aequales. Sit axis T14526789, et 12 sit abscissa, 23 ordinata; quaeritur tangens 3T. Dico fore T2 ad 23 ut 23/34 + 23/35 + 23/36 + 23/37 + 23/38 + 23/39 est ad 24/34 + 25/35 + 26/36 + 27/37 + 28/38 + 29/39 [fractions illegible in Dutens; restored from the differential of the given condition, the terms of the second sum taking + or − as the fixed point lies on one side of 2 or the other]. Eademque erit regula, continuatis tantum terminis, si non sex, sed decem vel plura puncta fixa supponerentur; qualia secundum methodos tangentium editas calculo praestare, sublatis irrationalibus, taediosissimae et aliquando insuperabilis operae foret: ut si rectangula plana vel solida secundum omnes biniones vel terniones possibiles ex rectis illis composita datae quantitati aequari deberent. In quibus omnibus, et multo implicatioribus, methodi nostrae eadem est opera, multo major, rarissimique exempli, facilitas. Et haec quidem initia sunt tantum Geometriae cujusdam multo sublimioris, ad difficillima et pulcherrima quaeque etiam mixtae Matheseos problemata pertingentis, quae sine calculo nostro differentiali, aut simili, non temere quisquam pari facilitate tractabit.
I shall show this by yet another example. Let the curve 133 be of such a nature that from any point of it, such as 3, the six lines 34, 35, 36, 37, 38, 39 drawn to six fixed points 4, 5, 6, 7, 8, 9 on the axis, added together, equal the given line g. Let the axis be T14526789, let 12 be the abscissa and 23 the ordinate; the tangent 3T is sought. I say that T2 will be to 23 as 23/34 + 23/35 + 23/36 + 23/37 + 23/38 + 23/39 is to 24/34 + 25/35 + 26/36 + 27/37 + 28/38 + 29/39 [fractions illegible in Dutens; restored from the differential of the given condition, the terms of the second sum taking + or − according as the fixed point lies on one side of 2 or the other]. And the rule will be the same, merely continuing the terms, if not six but ten or more fixed points were supposed; to do which by calculation according to the published methods of tangents, after removing the irrationals, would be a most tedious and sometimes insuperable labour, as it would be if plane or solid rectangles formed from those lines according to all possible combinations two by two or three by three had to be made equal to a given quantity. In all these cases, and far more complicated ones, our method requires the same work; its ease is much greater and of a most rare kind. And these are only the beginnings of a much more sublime Geometry, reaching to the most difficult and most beautiful problems even of mixed mathematics, which no one will readily handle with equal ease without our differential calculus or one like it.
11
Appendicis loco placet adjicere solutionem Problematis quod Cartesius a Beaunio sibi propositum, Tom. 3 Epist., tentavit, sed non solvit. Lineam invenire WW talis naturae, ut ducta ad axem tangente WC, sit XC semper aequalis eidem rectae constanti a. Jam XW seu w ad XC seu a, ut dw ad dx. Ergo si dx (quae assumi potest pro arbitrio) assumatur constans, sive semper eadem, nempe b, seu si ipsae x sive AX crescant uniformiter, fiet w aequ. (a/b) dw, quae erunt ipsae w ordinatae ipsis dw, suis incrementis sive differentiis, proportionales; hoc est, si x sint progressionis arithmeticae, erunt w progressionis Geometricae; seu si w sint numeri, x erunt logarithmi: linea ergo WW logarithmica est.
By way of appendix I should like to add the solution of the problem that Descartes, to whom it had been proposed by Debeaune, attempted in volume 3 of his Letters but did not solve: to find a line WW of such a nature that, when the tangent WC is drawn to the axis, XC is always equal to the same constant line a. Now XW, or w, is to XC, or a, as dw is to dx. Therefore if dx (which may be taken at will) is taken constant, that is, always the same, namely b, or if the x's, the AX's, increase uniformly, then w = (a/b) dw; so the ordinates w will be proportional to their own increments or differences dw. That is, if the x's are in arithmetic progression, the w's will be in geometric progression; or, if the w's are numbers, the x's will be logarithms. The line WW is therefore the logarithmic curve.
Notes.
- What this text is. The Nova methodus of October 1684 is the first printed statement of the differential calculus: the notation dx, the rules for sums, products, quotients, powers and roots, the conditions dv = 0 for an extremum and ddv = 0 for a point of inflection, and three applications. It appeared three years before Newton's Principia and twenty years before Newton published his fluxional calculus, and it became Exhibit A in the priority dispute of 1711–1716, conducted by the Royal Society. Leibniz was still writing to Remond (texts 70–73) and to Conti about that dispute when he died.
- "Algorithm" and the infinitangular polygon. Leibniz names his rules an algorithmus and defines the tangent as "the extended side of the infinitangular polygon which for us is equivalent to the curve". The calculus is presented as a blind symbolic procedure, the same ideal of a calculus ratiocinator that runs from the De arte combinatoria (1666) to the binary arithmetic of text 67: correct manipulation of characters yields truths without picturing them.
- Optics and least resistance. The refraction example derives the sine law from a minimum principle, the "easiest path", and refers back to Leibniz's Unicum opticae, catoptricae et dioptricae principium (Acta 1682). The principle that nature acts by the easiest ways is the physical face of the principle of the best that governs the metaphysical texts (62, 64, 65); Leibniz makes this link explicit in the Tentamen anagogicum (c. 1696).
- Debeaune's problem. Florimond Debeaune proposed this "inverse tangent" problem to Descartes in 1638; Descartes's attempt is in Clerselier's Lettres, vol. 3. Its solution, the logarithmic curve, is among the earliest printed solutions of an inverse-tangent problem by the new calculus.
- Text. The formulas in Dutens are badly garbled in the scan, and the 1684 printing itself contains misprints that were already noted in the eighteenth century. The rules are given here in their standard and correct form; the large example and the six-point tangent are reconstructed as marked. Gerhardt (GM V) has the best text and should be used for any check against the original typography.