You are on page 1of 4

Mathematics Magazine

ISSN: 0025-570X (Print) 1930-0980 (Online) Journal homepage: http://www.tandfonline.com/loi/umma20

The Sufficient Condition for the Differentiability of


Functions of Several Variables

Xu Pingya

To cite this article: Xu Pingya (1998) The Sufficient Condition for the Differentiability
of Functions of Several Variables, Mathematics Magazine, 71:4, 297-299, DOI:
10.1080/0025570X.1998.11996656

To link to this article: https://doi.org/10.1080/0025570X.1998.11996656

Published online: 11 Apr 2018.

Submit your article to this journal

View related articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=umma20
VOL. 71, NO. 4, OCTOBER 1998 297

The Sufficient Condition for the Differentiability


of Functions of Several Variables
XU PINGYA
Nanjing College of Electric Power
Nanjing210013
People's Republic of China

In courses on several variable calculus, the following observations about differentia-


bility are made. Given a function J: IR 2 ~ IR, it is only a necessary condition for
differentiability at the point P that both partial delivatives exist at P. It has been
observed in [1], [2], and [3] that if both partial delivatives exist and one of them is
continuous at the point P, then f is differentiable at P. An example given in [3] shows
that the corresponding assertion is not generally true for a function of n variables
when n ::=::: 3. Sp~cifically, for n ::=::: 3, the function need not be differentiable at the
point even if all n partial delivatives exist and one of them is continuous at the point.
We will prove the following theorem.

THEOREM 1. Suppose that the function J: IR n ~ IR has the property that all n partial
derivatives exist at the point P in IR n and n - 1 of the partial derivatives are
continuous at P. Then f is differentiable at P.

After the proof we provide an example to show that the conclusion need not hold if
two or more of the partial delivatives are discontinuous at P.

Proof Consider P = (x 1 , ... , x,.) and suppose that all the partial delivatives j;' exist
at P and that all of them are continuous at P except possibly f~. Then

and so

where lim~x ...... 0 e,. = 0. By Lagrange's mean value theorem,

where g1 is between x1 and x 1 + llx 1 for i = 1, 2, ... , n- 1.


2 2
Let p= V(llx 1 ) + ··· +(llxn) . Since f: is continuous at P, we have
298 MATHEMATICS MAGAZINE

so that

where limP .... 0 e; = 0 for i = 1, 2, ... , n - 1. Hence

and therefore

n
= I: (j( x1 , ... , X;_ 1 , X;+ Llx;, ... , Xn + Llxn)
i=l

n n n
L:J:(P)Llx 1 + L:e1Llx 1 = L:.t;'(P)Llx;+o(p).
i=l i=l i=l

This shows that f is differentiable at P and completes the proof of the theorem.
Now, for each n ~ 2, we give an example showing that the theorem no longer
remains true if more than one of the partials is allowed to be discontinuous. For
n = 2, many readers will find the example familiar. Let

if xr +X~= 0.
At the origin, we have

f 1'(O , ... , 0) = li m f(Lix,O,


A
... ,O) =0·,
Ax--+0 u. X

for xf + x~ * 0,

for each k. Thus f{ is discontinuous at the origin. Similarly, f~ is discontinuous at the


origin. Straightforward calculations show that the remaining partial derivatives are
continuous at the origin. When the point (Llx 1 , Llx 2 , ... , Llxn) tends to the origin
VOL. 71, NO. 4, OCTOBER 1998 299
along the line x 1 = x2 = "' = xn, we see that

This shows that function f is not differentiable at the origin.

REFERENCES

1. Zhang Jingqing, Wang Bowen, Xu Zhongyi, Tang Zhengyi, Wu Lingwen, and Chen Bisheng, Problems
Analysis of Calculus (Part B), Jiangsu Science and Technology Publishing House, Jiangsu, China, 1983.
2. J. M. Henle, Tangent planes with infinitesimals, Amer. Math. Monthly 91 (1984), 433-435.
3. Li Siyuan, A note on the differentiability of functions of several variables, Higher Mathematics 3 (1987),
21-23.

Math Bite: Equality of Limits in Ratio and Root Tests

Relations among various tests for convergence of series arose in the note [1] and its
corrigendum. For the convergence of series E::~o an with positive terms, two well-
known tests are as follows:

D'ALEMBERT'S RATIO TEST. Suppose that lim 11 -+oo an+l


an
= L. Then Lan conve~ges if
L < 1 and diverges if L > 1.
CAUCHY's ROOT TEST. Suppose that lim 11
--. "'a~! n = M. Then L:an converges if M <1
and diverges if M > 1.
We show-using the tests themselves-that if the limits Land M exist, they must
be equal. To this end, suppose that L < M. (The argument for the case M < L is
similar.) Choose a real number k such that L < k < M.
Now consider the series L:bn, where b" = a,JP. Then we have
bn+l
lim -b- =
L
k < 1, but lim b~ln = M > 1.
n-+oo n n->oo k
The first limit implies that L:bn converges; the second, that L:bn diverges. This is a
contradiction.

REFERENCES

1. D. Cruz-Uribe, SFO, The relation between the root and ratio tests, this MAGAZINE 70 (1997), 214-215;
corrigendum, this MAGAZINE 70 (1997), 310-311.

- PREM N. BAJA)
WICHITA STATE UNIVERSITY
WICHITA, KS 67260-0033

You might also like