The Relation of Differential and Delta Increments Author(s): C. G.

Phipps Reviewed work(s): Source: The American Mathematical Monthly, Vol. 59, No. 6 (Jun. - Jul., 1952), pp. 395-398 Published by: Mathematical Association of America Stable URL: http://www.jstor.org/stable/2306815 . Accessed: 29/02/2012 07:58
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp 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.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to The American Mathematical Monthly.

http://www.jstor.org

1952]

CLASSROOM NOTES

395

the differential calculus. In this case the "dx" in the integralis used to make the "x" a bound variable. It serves the same purpose that the first"j" does in the expression "Z>=3ki." (Notice that if we omitted the j and instead wroteZ5ki then we would be in doubt as to whether3i+4i+5i or k3+k4+k5 was intended.) While &f(x) would not be a good notation forthe integral (since x appears in thisexpressionas a freevariable), thereis no logical objection to a notationsuch as f,f. However, it would be difficult express l(x2-5)dx in this latter notato tion since thereis no commonlyused notationwhichdenotes the function whose value at x is x2-5. This functioncan be denoted by an expression such as XX(X2-5) [see The Calculi of Lambda-Conversiont], and if this notation were used forthe functionthen we could denote our integralby f:XX(x2-5) instead This concludes the discussion of our answer of "no" to (ii). of f(X2-5)dx. We now examine the circumstancesunder which the answer to (ii) is "yes." Let us suppose thatf is a continuousfunction. Then it is possible to prove that thereexist functions such that g' =f. It followsfrom(i) that dg=f. dl forsuch g functions We may now defineff dI to be the class of all functions forwhich g. g dg=f dl. (The "dl" inf dI is actually superfluous since it is just as reasonable f to denote this class of functions ff.) Finally,we can define f dl = g(b) - g(a) by where g is a memberof the class fd dl. Thus, if integrationis introducedvia the indefinite integralor primitivemethod,and "dx" representsthe differential of the identityfunction,then it is possible to interpretthe "dx" in f:f(x)dx as being the same as the "dx" of the differential calculus. 5. Conclusion. The author feels that most of the confusionconcerningdifferentials disappears if the instructor: (1) uses the word "variable" only to mean a symbol (i.e. an ink spot or chalk mark,etc.,) and avoids the usual implicationthat a variable is a mysterious Dr. Jekylland Mr. Hyde type of gadget; (2) carefullydistinguishesbetween a function and the numberf(x) which f is the value off at the numberx; of of and not the differentials variables. (3) definesthe differentials functions
THE RELATION OF DIFFERENTIAL AND DELTA INCREMENTS
C. G. PHIPPS, Universityof Florida

the use Do mathematiciansin general have as much trouble understanding notation as is indicated by the sporadic referand meaning of the differential ences to it which have appeared in this MONTHLY? It is to be hoped not, for is defined. thereshould be no confusionwhenthesymbolism properly This note is writtenin response to the Editor's two queries: (1) what is a notation related to that forintegraand (2) how is the differential differential is tion? One consistentset of definitions given here whichanswers the questions in the orderthey are proposed.
The CalculiofLambda-Conversion, Princeton Press,1941. University t AlonzoChurch,

396

CLASSROONI

NOTES

[June

Let y=F(x) be a continuous functionwith continuous derivatives of all orders. The derivative of this functionis found by the usual limitingprocess and the resultmay be indicated by any one of the usual symbols,say Dry. Let (x, y) and (x', y') be two different solutions of the above equation. We definethe delta-increments as (la) zxx= x'-x and Ay= y'-y. Then we definethe differentials these variables as of (lb) dx = Ax and dy = (Dxy)Ax.

The first definition (lb) applies specifically the independentvariable and in to these definitions consistent are the second to the dependent one. Furthermore, for,ifwe writey = x, the second definition yields the same resultwhichis stated in the first. From the latter two definitions obtain the usual forms we (2a) dy = (Dxy)dx and (2b)
-=

dx

Dxy.

The second of these formssupplies the reason forwritingthe derivative in its usual fractional form. also showsthechange fromn to dx gives a symmetry It Ax to the notation. The first these formshas two interpretations: of (i) it is a rule for carryingout the operation of "taking the differential"; and of (ii) it is a formula for computing the value of the differential the dependent variable. The rightmemberof (2a) is a function two independentquantities,x and of dx. If dx is held constant,dy becomes a functionof x alone to which we can apof ply the rule in (i) and definethe second differential y as (3a) d(dy) = d'y
=

[(D'y)d x]dx = (D y)dx .

Further,we definethe higherdifferentials y as the resultof successive appliof cations of this rule underthe same conditionsand obtain
(3b) dny = (D7y)dx7.

into the usual fractional These differential equations may then be transformed notation forderivativesof higherorders. If the functionF(x) is expanded into a Taylor's series,we may write (4a) Ay = (Dxy)dx + -2 (D y)dx2+ *

21

(D'y)>dx' +

into this series may be transformed By the aid of our definitions

19521
(4b)

CLASSROOM NOTES

397

1 1 Ay = dy+-d2y+-d3y+ 2! 3!it

-d ny+.

Thus it is seen that our definitions convenientas well as consistent. are that withinthe intervalof convergencethe partial sums of It is well-known the terms of this series may be used as successive approximationsto Ay. For termalone gives the value Ay would have if its rate of change instance,the first two termsyield the value were constant while x changed fromx to x'. The first Ay would have if this rate of change itselfhad changed at a constant rate. The And partial firstthree termshave the constant rate removed one step further. interpretations. sums with more termshave corresponding If the functionF(x) does not possess continuousderivativesbeyond the n-th At order,we can of course obtain no more than the n-th differential. the same time,the seriesapproximationwould likewisenot extend beyond the n-thorder. of to The extensionof these definitions functions more than two variables is illustratedby two examples. Let z = G(x, y). As independentvariables, we deof Then finethe differentials x and y as equal to theirdelta-increments. (5) dz- Gdx + G,dy,

This formgives both a rule where the subscriptdenotes partial differentiation. and a formulaforcomputingits value. The definifor "taking the differential" the tion may be extended to any order. Furthermore, relation between Az and of its differentials all ordersis given by a formidenticalwith that in (4b). If z = G(x, y) where y = F(x), the value and formof dz is still given by (5) however, is whereinthe value of dy is given by (2a). The second differential, more complicated; it is (6) + + d2z = Gxxdx2 2Gxudxdy G,,dy2+ G,d2y. of morecomplicated but can The differentials higherordersare correspondingly withcare be writtendown. Also, the formin (4b) still holds true. Turning now to the notation for integration,we find its connection with notation dependent upon the meaningattached to the integral. differential the operatorde(a) We may begin by defining integralsign as the indefinite finedin the equation

f = y +C. dy

If y = F(x) and D. y =f(x), we have by substitutionthat

f(x)dx = F(x) + C.

integrationis defined(except forthe constant C) as the By this definition, inverse operation of "taking the differential." Hence, the integral sign as an

398

CLASSROOM NOTES

[June

operator in this sense has no meaning unless it is applied to a perfectdifferential. we (b) Alternately, may defineAy exactly by the expression AY=

f
xo

r0+x+a

f(x)dx,

wherey = F(x) and Dxy=f(x) as above. The value of this integral may be approximated by taking f(x) =f(xo) throughoutthe interval. This approximate value is (dy=)f(xo)Ax. A better approximationmay be obtained by dividingthe intervalinto more and smaller parts and using the above approximationforeach part. The limit as the numis ber of parts becomes infinite the exact value of the integral. This method of approximatingAy does not correspond,except forthe first step, with the method which uses eitherformof (4). Hence the dx in this inteof gral has no relationeitherto Ax or to the previousform dx. The only obvious reason forusing the same symbolismin (b) as in (a) is that the definitionin (a) may be used to evaluate the limit representedby (b), namely as F(xo + Ax)
-

F(xo).

the Thus by writingthe definiteintegral in the same formas the indefinite, As former. formmay be used to findthe value ofthe knownproperties thelatter of one gives the form,the other gives the subwith differentials, interpretation stance.

TOWARD UNDERSTANDING DIFFERENTIALS H. J. HAMILTON, PomonaCollege

toThe mathematicsteacher hardlyneeds to excite himself 1. Introduction. with "nascent" or "evanescent" increof day about identification differentials If ments or with "vanishinglysmall" quantities, or even with "infinitesimals." physics has not entirelyabandoned these mystical ideas, at least mathematics has done so. It is now standard mathematical pedagogy to introducedifferentials as follows.Let y=f(x), wherex is the independentvariable. Then we define dx_=Axand dy-f'(x)dx, where Ax is an arbitrary,non-zero incrementin x. A and we explain to our students that a perpicture goes with these definitions, to fectlygood value fordx is the distance in millimeters the remotest known star; dy/dxis still equal to f'(x). (Of course, if we are as carefulas we should be, we say that dx is a variable for increment, we need to "see" the "dx" in ratherthan an arbitrary increment where to give dx an arbitraryvalue would most cases-certainly in integrands,