Sunteți pe pagina 1din 4

The Differentiation of a Composite Function

Author(s): A. F. Beardon
Reviewed work(s):
Source: The Mathematical Gazette, Vol. 55, No. 394 (Dec., 1971), pp. 395-397
Published by: The Mathematical Association
Stable URL: http://www.jstor.org/stable/3612366 .
Accessed: 18/12/2012 00:52

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.

The Mathematical Association is collaborating with JSTOR to digitize, preserve and extend access to The
Mathematical Gazette.

http://www.jstor.org

This content downloaded on Tue, 18 Dec 2012 00:52:01 AM


All use subject to JSTOR Terms and Conditions
ESTABLISHING A PECKING ORDER 395

These are gathered: (n - 3) + 3(n -6) + 7(n - 12) + 15(n - 24) + ...
i.e.

[(n-l)+(n-2)+(n-3)+... +1]
-[(n - 2?) + (n - 21) + (n - 22)+ . .. + (n - 2r),]

where " r " is the exponent of 2 such that 2r < n < 2r+l. To this
is added the bonus gained in Congress (5), giving

[n(n 1)+ 1]-[(r+ l)n- (2r+1-1)]


or
(n - 2r - 3) + 2r+l bonuses in Congress (n)

and as there are n(n - 1)/2 pairings, the number of weighings


necessary and essential is
n(r+ 1)-2r+l.
This formula confirms empirical values for all n > 5. For
1 < n < 4, the values it generates must be increased by 1.

1 King EdwardClose, JOHN CAMERON


Hastings,
Sussex

THE DIFFERENTIATION OF A COMPOSITE FUNCTION

BY A. F. BEARDON

In a previous issue of the Gazette[2] J. G. Colman gives a formula


for the nth derivative of the composite function f[g(x)], states that
this is proved by induction and asks for a more direct and illumina-
ting proof. The formula is
n i[g(i)(x)]Ai d
Dnf[g(x)] = f(A +An)[g(x)]l fl D = (1)
i=x
Ai(i!)'
AEPn

where P, is the set of ordered n-tuples A = (A1,. . ., An) of non-


negative integers which satisfy
A1+2A2+ . . . nA = n.
We now give a direct proof in the case when f and g are analytic
functions and this shows quite clearly how the formula (1) arises.
The idea of this proof is very simple. We first express f(y) as a
c

This content downloaded on Tue, 18 Dec 2012 00:52:01 AM


All use subject to JSTOR Terms and Conditions
396 THE MATHEMATICAL GAZETTE

power series in (y-g(a)) and then substitute the power series for g
about x = a in place of y. This gives f[g(x)] as the composition of
two power series, the coefficients of which involve the derivatives of
f at g(a) and of g at a. The re-arrangementof this composition into a
single power series is valid in some neighbourhood of x = a (see [1],
p. 414) and the coefficients of this series must be the Taylor co-
efficients of f[g(x)] at x = a. Thus we can express the nth deri-
vative of f[g(x)] at x = a in terms of the derivatives of f and g, and
this yields (1).
We now give a formal proof. If g is analytic in some neighbour-
hood of a, then, in that neighbourhood,

(x)= g()(a) (x -a)'.


g(x) =S=O s

Similarly, if f is analytic in some neighbourhood of y = g(a), then


in that neighbourhood,
f(Y) =E Yr!
r=O
It follows that, for all x in some neighbourhood of x = a,

fig9(x)]= (ssEgs)(a)
Ef(r)[g(a)]
r! (x a)8)r
rO
It is easily seen that termwise differentiation is valid and so we
have
f_(r)[g(a]
Dnf[g(x)]= E r!)] Dn fl(5 ! (x a)
(2)(s)(a) - (2)

We shall put x = a in (2) and so we are interested in the coefficient of


(x - a)n in the expansion of

(s)
( xg (a) (X- a)s
Clearly this is the same as the coefficient of (x - a)n in the expansion
of
n g(s)(a)(x-a)s

The simplest way to find this is to use the Multinomial Theorem


([3], p. 111):
1
(Yi + ** *Y)r -- ! A ..
A .
nY

where the sum is taken over all finite sequences (A1,..., An)in which
the Aiare non-negative integers satisfying A1+ . . . + An= r. If we
now substitute
ys =g()(a)
8! (x-a)'(s =
=..s
1,...,n) n)

This content downloaded on Tue, 18 Dec 2012 00:52:01 AM


All use subject to JSTOR Terms and Conditions
THE DIFFERENTIATION OF A COMPOSITE FUNCTION 397

we find that the function (3) is

r!
E (x a)Al+2A2+... +nAn [g(i)(a)/i!]A,
A1;... An, i=l

Substituting x = a in each side of (2) we find that

Dnf[g(a)] = Zf(r
r=0 r!g(a)] .n! * Ai! fI [g(i)(a)/i!]Ai
A1!. . . An! i=1
where Z* denotes summation over those (A1,. .. , An) in which
Al + ... + An = r and A1+ 2A2 + ++nAn =n. In an obvious sense,
we have

_r=OAn)

and so (1) follows at x = a.

REFERENCES
1. T. M. Apostol, MathematicalAnalysis, Addison-Wesley,1957.
2. J. G. Colman, The differentiation of a function of a function,
Math. Gaz., LIV, No. 390, p. 389 (1970).
3. J. G. Kemeny, J. L. Snell and G. L. Thompson, Introductionto
finite mathematics, Prentice-Hall, 1966 (2nd edition).

St. Catharine's College, A. F. BEARDON


Cambridge

SOME PROJECTIONS OF THE HYPERCUBE 74

BY G. M. HAMILTON

It is possible [1] to project the four-dimensionalhypercube y4 into


three dimensions in a symmetrical manner so that the edges lie
mainly along the edges of a rhombi-dodecahedronbut with two of the
vertices coincident at the centre of the figure. The length of the

original sides of the unit hypercube are foreshortened by 2-. It is


also possible to project y4 into two dimensions as an octagon (Figure
1) with eight of the vertices disposed symmetrically in the middle of
the figure. The question arises as to what is the configurationof the
intermediate three-dimensional projection of the hypercube which
gives Figure 1 on a second projection. Clearly the symmetrical

This content downloaded on Tue, 18 Dec 2012 00:52:01 AM


All use subject to JSTOR Terms and Conditions