Paper / 论文 ·
On the Generality of General Calculus
In On Recondite Geometry, Leibniz boldly claims that ‘while before now algebraists have taken letters or general numbers for the quantities sought, I have taken general or indefinite equations for the lines sought in such transcendent problems’. The force of his calculus, under this context, is viewed to rest not simply on a higher, more abstract generality of quantities, but a generality of a different kind. As algebra is an abstraction from definite quantities, so is calculus an abstraction from definite equations. Since any definite algebraic equation expresses the law of correlation between indefinite quantities, Leibniz interprets calculus to be an abstraction from definite laws, a principle of all principles.
In order to substantiate this claim, we have to retrace the steps in this unfolding of generality within calculus. What are these generalities? To what problems do these generalities pertain? Every abstraction is effected to resolve a problem, i.e. an exception to the rule, that calls for a more general determination of a solution. I hereby claim without proof: to every derived generality corresponds an original exception. This should be the axiom for any mathematical construct.
The exception Leibniz is confronted with in On Recondite Geometry are transcendent quantities. From the start of the main body of this text, he reveals the problem by revealing one ‘source of Transcendent Quantities, that is, a reason why some problems are neither plane, nor solid, nor sursolid, nor of any definite degree, but transcend every algebraic equation’. What are the problems? Leibniz generously provides two: ‘cutting an angle’ and ‘finding an arbitrary number of mean proportionals’. These are problems because they vex the limits of preceding algebraic geometry: they, in any of our attempts to solve them, transcend definite degrees. The resources of preceding algebraic geometry cannot cope with such indefiniteness.
Yet what constitutes a problem? What is the reason for our determination of some things being problems, while others not? In the same passage, Leibniz gives the following reasons for including transcendent quantities into ‘the general problematic’ of geometry: 1. proposable as a problem inside geometry, that is, the problem should be one that naturally arises from a geometrical operation, 2. determinate, 3. describable by an exact and continuous motion, 4. useful.
From such determinations of what a geometrical problem is we should be capable of deriving the requisite characteristics of a mathematical generality that is devised to solve these problems. First, the generality should still be immanent to the original scope. For instance, a metaphysical generality cannot be either sufficient or desirable as a solution to a geometrical problem. Second, the generality should provide a generative principle that answers to any specific instance of determination. Third, this generality should be inherently unified. This requirement ensures that generality is not a simple grouping of ad hoc premises or rules, but must be one rule universally applicable to any specific problem. Otherwise, the relationship between one problem to another, or in Leibniz’s own terms, the ‘motion’ from one instance to another, like from one point to another point on the same curve, would be neither continuous nor exact. Finally, the generality should be simple, so that the solution is practical.
Leibniz’s stroke of genius here, however, may consist in this alone: instead of moving between a proposed generality and specific problems, he identifies the problem to be the generality itself. A general rule is either inferred or deduced. An inferred generality can only incompletely solve the problems, since new exceptions might spring up and refute the rule previously instituted. But if all such specific problems are grasped as one general problem, the deduced general rule necessarily solves all of them all at once. Therefore, in the following paragraph, Leibniz shifts his focus from the specific problems of transcendent quantities to the problem of transcendence in general: ‘moreover, because the method of investigating indefinite quadratures of their impossibilities is for me only a special case (and indeed an easier one) of a much greater problem, which I call the inverse Method of Tangents, in which the greatest part of transcendent geometry is contained, and because if this problem can be solved algebraically, all sorts of discoveries might be made...’
The general problem formulated by Leibniz is: given any property of the relation between two quantities, one is to find the curve that expresses this relation. The given property can be in any form. The examples Leibniz gives in his On Recondite Geometry are illuminating. It could be a property relating the rate of change of one quantity to the other quantity (dx/dy). The problem is in this case called an ‘inverse tangent problem’. It could also describe the positioning of two quantities as in one geometrical figure: one quantity being the abscissa on the radius of a circle, while the other being the arc determined in relation to this abscissa. Despite apparent differences, the latter example can eventually be reduced to an inverse tangent problem. For if we relate the rate of change of the arc length to the abscissa of the given circle, we discover a fixed relationship between the two through similar triangles. Therefore, this specific geometrical problem of representing arc length through its abscissa is transformed into a special case of inverse tangent problems. The same holds for problems of quadrature, since any problem of quadrature is solved through establishing the rule for the tangents of the quadratrix. For the area under the quadranda is obtained through integrating the infinitesimal rectangles formed by the ordinates and the differences in abscissa (∫ydx). The rate of change of the ordinate of the quadratrix in relation to its abscissa is therefore determined as the rate of change of the area under the quadranda in relation to the abscissa (dv = d∫ydx = ydx), which yields the tangent of the quadratrix to be the function of the quadranda (dv/dx = y). All these specific geometrical problems, therefore, can be reduced to the general problem of inverse tangents. The specific problems of transcendent curves can accordingly be reduced to one general problem of transcendence, since this general problem of transcendence is itself only a special case of the inverse tangent problem.
To grasp transcendent problems as one problem of transcendence does not mean, however, to simply identify the common characteristic among these transcendent quantities, though it might be the preliminary step. Instead, what is crucial in such metamorphosis is to establish the proper way to represent this common characteristic mathematically. The Leibnizian expression of this general problem is formulated thus:
0 = a + bx + cy + exy + fx2 + gy2 etc. + v
in which ‘v’ stands for the transcendent quantity. Its transcendence is not simply stipulated, but also manifest in contrast to what precedes, i.e. an indefinite algebraic expression of quantities of finite degrees. Moreover, transcendence in general is only intelligible in such conjunction to finite quantities: ‘v’ by itself only represents an indefinite quantity that is not exclusively a transcendent one, whereas symbols such as ‘vw’ or ‘cos v’ are too specific for transcendence in general. Therefore, one indeed can say the Leibnizian expression of a general transcendence is its most proper expression.
The other important aspect of this formula is already hinted at at the beginning of this essay. Leibniz claims that his originality here consists in taking ‘general or indefinite equations for the lines sought in such transcendent problems’. What he is effectively saying is that, he reinvents the ‘general problematic’. Every generality is devised against a set of problems. A genus is posited to include and help define all the related species. What is unique about algebra and the analytical tradition it continues, however, is that the generality is transferred from the realm of solutions to the realm of problems. Instead of positing the generality as an answer to the set of problems, the generality itself becomes the problem. In other words, ‘x’, which stands for the general number, is not a solution to a problem, like a genus is to the definition of a species, but a symbol that represents that which is to be sought. The algebraic symbol stands for something unknown and is thereby posited as a problem, whereas the metaphysical genus is a known generality proposed to resolve problems of particular definitions.
Leibniz inherits from this tradition of thinking. For him, the generality of calculus as a modus operandi is first and foremost founded on the generality of the problem it encounters. The curve, whether transcendent or algebraic, needs to be first posed as a ‘general’ or ‘indefinite’ equation that is to be solved by a general method. The generality of calculus, therefore, rests upon confronting not a set of particular problems, e.g. cutting angles and finding mean proportionals, but rather the one general problematic of finding any curve whatsoever, i.e. the inverse tangent problem. It is only due to this fact that Leibniz objects to the prevalent misconception on the part of other mathematicians: ‘sometimes some things remain arbitrary, in which case innumerable lines can be found that satisfy the problem, which is the reason why many, seeing it afterwards, might think that the problem is not sufficiently defined and is not in our power’. The fault of others, according to Leibniz, consists not so much in denying the general nature of the solution, i.e. the method through which ‘innumerable lines can be found’, but primarily in rejecting the possible generality of a problem. To their own disadvantage, they wrongly believe that only the method for arriving at particular solutions for particular problems is general and do not realize that the problem itself can also be generalized. The true Leibnizian renovation is only intelligible in this context of a tradition of general problematic. What he invents majorly hinges on his widening the scope of what counts as a problem. He surpasses preceding algebraists mainly in this sense, that the general problem for preceding algebra is the unknown quantity, whereas the general problem for calculus becomes the unknown relation between quantities. A function is that which expresses the relation between quantities. More specifically, it expresses the relation between the ordinates and abscissas of a geometrical figure. To solve the general problem of finding curves is thus to solve the general problem of establishing laws between quantities.
The internal structure of On Recondite Geometry can therefore be revealed via the following logical steps that Leibniz takes in order. First, Leibniz claims that the specific problems that involve transcendent quantities can be reduced to one general problem of inverse tangents. Then he gives a mathematical expression for the general problem of inverse tangents: given any tangent property, one can solve for the curve through employing the algebraic expression of any indefinite and general equation; if it turns out unsolvable, one can be certain that the problem is a transcendent problem. Next, he provides multiple examples of solving these transcendent problems with the implication that these problems can be reduced to the general problem through his differential and integral calculus.
Finally, we should determine if the generality of calculus fits the criteria we derive in the earlier part of this essay. To be sure, when Leibniz posits those characteristics of geometrical problems he has a more limited and specific set of problems in mind. But the criteria should not vary no matter on what scale they are applied. Here I restate the criteria for what constitutes a geometrical problem: 1. inherence within geometry; 2. determinateness; 3. describability through continuous and exact motion; 4. usefulness. The corresponding criteria which we derive for a generality that ‘solves’ such problems are: 1. the same inherence within geometry; 2. universal generativeness; 3. unity; 4. simplicity.
Now, it is not difficult to see that the generality of method of differential calculus depends upon the generality of the problem it posits for itself. Since integration is only the reciprocal process for differentiation, and for any inverse tangent problem, including transcendent problems, the ratio between differential quantities (tangent) is determinate, all inverse tangent problems are solvable through integration. Hence so long as the problem of inverse tangents includes all geometrical problems, the method of calculus is the general method for any problem. As Leibniz himself claims: ‘hardly anything can be imagined that is more useful, concise, and universal for treating transcendent problems by calculation wherever measurements and tangents occur than my differential calculus or analysis of indivisibles and infinites’. Therefore the question now becomes: is the problem of inverse tangents general enough?
I claim this problem of inverse tangents is properly considered the general problem for Leibnizian calculus. This problem of inverse tangents, qua problem, arises naturally within the pursuit of geometrical determination and is determinate, though not particular. It also describes a space in which a curve is continuously varied and modulated so as to ‘satisfy the problem’, such as all logarithmic lines are lines which conform to Barrow’s Theorem. Qua generality, this general problem universally represents all possible determinations of specific problems and itself maintains one simple expression, i.e. through the mathematical formula of an indefinite equation as the unknown. Therefore, the problem calculus defines for itself is properly general. The only reason why one might hesitate on this issue is that one cannot be sure if all specific geometrical problems can be reduced to a problem of finding the curve for two related quantities, and if it is always possible to determine the relationship between the differentials of these two quantities, since these are the prerequisites for transforming a geometrical problem into an inverse tangent problem. Yet I believe it is sufficient to reply that only a problem that can be turned into a inverse tangent problem constitutes a proper geometrical problem. A problem where the differential quantities are not determinable contradicts the criterion of determinateness and continuity (for if the differentials of two quantities do not maintain one singular relationship, then either the two quantities are totally indeterminate with respect to each other, or as at a certain point their relationship shall change, they are not continuous in relation to each other), and is therefore not a geometrical problem at all. The problem of inverse tangents is therefore the most general problem of geometry, and calculus the highest generality.
本文由 AI 从英文原文翻译。请以原文为准。
在《论深奥的几何》中,莱布尼茨大胆宣称:“此前,代数学家用字母或一般的数表示所求的量,而我则用一般的或不定的方程,表示这类超越问题中所求的线。”在这一语境中,他的演算的力量,并非仅仅建立在量的更高、更抽象的普遍性之上,而是建立在一种不同种类的普遍性之上。正如代数从确定的量中进行抽象,微积分则从确定的方程中进行抽象。既然任何确定的代数方程都表达了不定的量之间相互关联的法则,莱布尼茨便把微积分理解为从确定法则中进行的抽象,是一切原则的原则。
为了使这一主张得到充实,我们必须重新追踪微积分内部普遍性展开的步骤。这些普遍性是什么?它们关联着什么问题?每一次抽象都是为了处理一个问题,也就是规则的一个例外,而这一问题要求对解答作出更普遍的规定。我在此不加证明地断言:每一个派生出来的普遍性,都对应着一个原初的例外。这应该是任何数学构造的公理。
莱布尼茨在《论深奥的几何》中面对的例外,是超越量。从正文一开始,他就通过揭示一个“超越量的源头,也就是某些问题既非平面问题,也非立体问题、超立体问题或任何确定次数的问题,而是超出一切代数方程的一个原因”,来揭示问题。这些问题是什么?莱布尼茨慷慨地提供了两个:“分割一个角”与“寻找任意数目的比例中项”。它们之所以是问题,是因为它们逼迫此前的代数几何显露其界限:在我们试图求解它们的任何尝试中,它们都超出确定的次数。此前代数几何的资源无法应付这样的不定性。
然而,什么才构成一个问题?我们凭什么判定某些事情是问题,而另一些不是?在同一段落中,莱布尼茨给出了将超越量纳入几何的“普遍问题域”的如下理由:1. 它能够作为问题在几何内部被提出,也就是说,这个问题应当从一种几何操作中自然产生;2. 它是确定的;3. 它能够由精确且连续的运动来描述;4. 它是有用的。
从这些关于几何问题是什么的规定中,我们应当能够推导出,为解决这些问题而设计的数学普遍性应具有什么必要特征。首先,普遍性应当仍然内在于原来的范围。例如,一种形而上学的普遍性,既不能充分解决一个几何问题,也不是我们想要的解答。其次,普遍性应当提供一个生成原则,回应任何特定的规定实例。再次,这一普遍性应当内在地统一。这个要求确保普遍性不是若干权宜前提或规则的简单集合,而必须是一条普遍适用于任何特定问题的规则。否则,一个问题与另一个问题之间的关系,或者用莱布尼茨自己的话说,从一个实例到另一个实例的“运动”,比如从同一曲线上的一点到另一点的运动,就既不连续,也不精确。最后,普遍性应当是简单的,从而使解答能够被实际运用。
然而,莱布尼茨在这里的天才之举,也许仅仅在于这一点:他并非在一个被提出的普遍性与特殊问题之间来回移动,而是认定问题就是普遍性本身。一个普遍规则要么通过归纳推得,要么通过演绎得出。归纳而来的普遍性只能不完全地解决问题,因为新的例外可能涌现,推翻先前确立的规则。但如果所有这类特殊问题都被把握为一个普遍问题,那么演绎得出的普遍规则就必然一次性地解决所有问题。因此,在接下来的段落中,莱布尼茨把重点从超越量的特殊问题,转移到普遍的超越性问题上:“此外,因为对我来说,探究不定求积或其不可能性的方法,只是一个大得多的问题的特例(而且还是较容易的特例);这个问题我称为逆切线法,超越几何的最大部分都包含其中;又因为如果这个问题能够用代数求解,就可能作出各种发现……”
莱布尼茨表述的普遍问题是:给定两个量之间关系的任意性质,求表达这一关系的曲线。所给定的性质可以采取任意形式。他在《论深奥的几何》中给出的例子很有启发性。它可以是一种把一个量的变化率与另一个量联系起来的性质(dx/dy)。在这种情况下,问题被称为“逆切线问题”。它也可以描述同一几何图形内两个量的位置:一个量是圆的半径上的横坐标,另一个量则是相对于这一横坐标确定的弧。尽管表面上有所不同,后一个例子最终也可以还原为逆切线问题。因为,如果我们把弧长的变化率与给定圆的横坐标联系起来,就会通过相似三角形发现两者之间的固定关系。因此,这个通过横坐标表示弧长的特殊几何问题,被转化为逆切线问题的一个特例。求积问题也同样如此,因为任何求积问题都是通过确立求积曲线的切线法则来解决的。因为待求积曲线下的面积,可以通过对纵坐标与横坐标的差所构成的无穷小矩形进行积分而得到(∫ydx)。因此,求积曲线纵坐标相对于其横坐标的变化率,就被确定为待求积曲线下的面积相对于横坐标的变化率(dv = d∫ydx = ydx),由此得到,求积曲线的切线是待求积曲线的函数(dv/dx = y)。因此,所有这些特殊的几何问题,都可以还原为逆切线这一普遍问题。超越曲线的特殊问题也相应地可以还原为一个普遍的超越性问题,因为这一普遍的超越性问题,本身也只是逆切线问题的一个特例。
然而,把超越问题把握为一个超越性问题,并不意味着仅仅辨认这些超越量所共有的特征,尽管这可能是预备步骤。相反,这一变形中关键的,是确立以数学方式表示这一共同特征的恰当方式。莱布尼茨将这一普遍问题表达为:
0 = a + bx + cy + exy + fx2 + gy2 等等 + v
其中,“v”代表超越量。它的超越性并非仅仅被规定下来,而是通过与前面部分——即由有限次数的量组成的不定代数表达式——的对照而显现。此外,普遍的超越性,只有在与有限量的这种结合中才是可理解的:“v”本身只表示一个不定的量,并不专门是超越量;而诸如“vw”或“cos v”这样的符号,对于普遍的超越性来说又过于特殊。因此,我们确实可以说,莱布尼茨对一般超越性的这一表达,就是最恰当的表达。
这一公式的另一个重要方面,在本文开头就已经有所暗示。莱布尼茨宣称,他在这里的原创性在于用“一般的或不定的方程,表示这类超越问题中所求的线”。他实际上是在说,他重新发明了“普遍问题域”。每一种普遍性都是针对一组问题设计的。一个属被设定出来,是为了包含并帮助定义所有相关的种。然而,代数及其所延续的分析传统的独特之处,在于普遍性从解答的领域转移到了问题的领域。普遍性不再被设定为一组问题的答案,而是自身成为问题。换句话说,代表一般的数的“x”,并非像属之于种的定义那样,是一个问题的解答,而是表示所求之物的符号。代数符号代表某个未知的东西,因而被设定为问题;而形而上学的属,则是一种已知的普遍性,被提出来解决特殊定义的问题。
莱布尼茨继承了这一思考传统。对他而言,微积分作为一种操作方式的普遍性,首先且最根本地奠基于它所面对问题的普遍性。曲线,无论是超越曲线还是代数曲线,都需要首先被提出为一个有待以普遍方法求解的“一般的”或“不定的”方程。因此,微积分的普遍性所依赖的,不是面对一组特殊问题,比如分角与寻找比例中项,而是面对寻找任意曲线的一个普遍问题域,即逆切线问题。也正因为这一点,莱布尼茨反对其他数学家普遍存在的误解:“有时某些东西仍然是任意的,在这种情况下,可以找到无数条满足该问题的线;这就是为什么许多人事后看到它时,可能认为问题没有得到充分规定,并且不在我们的能力范围之内。”按照莱布尼茨的看法,他人的错误与其说在于否认解答的普遍性质,即否认“可以找到无数条线”的方法,不如说首先在于拒绝问题可能具有的普遍性。他们错误地相信,只有为特殊问题寻求特殊解答的方法才是普遍的,却没有认识到,问题本身也可以被普遍化,这使他们自己处于不利之地。莱布尼茨真正的革新,只有在这一普遍问题域传统的语境中才是可理解的。他的发明主要在于扩展了什么可以算作问题的范围。他主要在这一意义上超越了此前的代数学家:此前代数的普遍问题是未知的量,而微积分的普遍问题则成为量之间未知的关系。函数是表达量之间关系的东西。更具体地说,它表达几何图形的纵坐标与横坐标之间的关系。因此,解决寻找曲线的普遍问题,也就是解决在量之间确立法则的普遍问题。
因此,《论深奥的几何》的内在结构,可以通过莱布尼茨依次采取的以下逻辑步骤揭示出来。首先,他主张,涉及超越量的特殊问题可以还原为一个普遍的逆切线问题。然后,他为逆切线这一普遍问题给出数学表达:给定任意切线性质,就可以运用任何不定的一般方程的代数表达式,求解相应曲线;如果结果无法求解,就可以确信,这个问题是一个超越问题。接下来,他给出多个解决这些超越问题的例子,并暗示这些问题可以通过他的微分与积分演算还原为那个普遍问题。
最后,我们应当确定,微积分的普遍性是否符合本文前面推导出的标准。诚然,当莱布尼茨提出几何问题的那些特征时,他心目中是一组更有限、更具体的问题。但无论应用于何种尺度,标准都不应有所变化。在此,我重述构成几何问题的标准:1. 内在于几何;2. 确定性;3. 能够由连续而精确的运动来描述;4. 有用性。而我们推导出的、“解决”这些问题的普遍性所对应的标准,则是:1. 同样内在于几何;2. 普遍的生成能力;3. 统一性;4. 简单性。
现在,不难看出,微分演算方法的普遍性,依赖于它为自己设定的问题的普遍性。既然积分只是微分的逆过程,而对于任何逆切线问题,包括超越问题,微分量之间的比(切线)都是确定的,那么所有逆切线问题都可以通过积分求解。因此,只要逆切线问题包含所有几何问题,微积分的方法就是解决任何问题的普遍方法。正如莱布尼茨自己宣称的:“凡是涉及度量与切线的地方,要通过计算处理超越问题,几乎无法想象有什么比我的微分演算,或不可分量与无限量的分析更有用、更简洁、更普遍。”因此,现在的问题变成:逆切线问题是否足够普遍?
我主张,应当恰当地把逆切线问题视为莱布尼茨演算的普遍问题。逆切线问题作为问题,自然地产生于对几何规定的追求之中,而且是确定的,尽管不是特殊的。它还描述了一个空间,在这个空间中,一条曲线被连续地变化和调节,以便“满足问题”,比如所有对数曲线都符合巴罗定理。作为普遍性,这个普遍问题普遍地表示了特殊问题的一切可能规定,并且自身维持着一个简单的表达,即通过作为未知者的不定方程这一数学公式来表达。因此,微积分为自己界定的问题是恰当地普遍的。人们对此可能有所迟疑的唯一理由,是无法确信所有特殊几何问题是否都可以还原为:为两个相关的量寻找曲线的问题;也无法确信是否总能确定这两个量的微分之间的关系,因为这些是把几何问题转化为逆切线问题的先决条件。然而,我认为这样回答就已足够:只有能够转化为逆切线问题的问题,才构成恰当的几何问题。一个微分量无法被确定的问题,与确定性和连续性的标准相矛盾(因为,如果两个量的微分不保持一个单一关系,那么,要么两量彼此完全不确定,要么它们的关系会在某一点改变,因而它们彼此并不连续),所以它根本就不是几何问题。因此,逆切线问题是几何中最普遍的问题,而微积分则是最高的普遍性。