Nearness--A Better Approach to Continuity and Limits

Author(s): P. Cameron, J. G. Hocking and S. A. Naimpally

Source: The American Mathematical Monthly, Vol. 81, No. 7 (Aug. - Sep., 1974), pp. 739-745
Published by: Taylor & Francis, Ltd. on behalf of the Mathematical Association of America
Stable URL: https://www.jstor.org/stable/2319561
Accessed: 25-09-2019 15:09 UTC

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.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Mathematical Association of America, Taylor & Francis, Ltd. are collaborating with
JSTOR to digitize, preserve and extend access to The American Mathematical Monthly

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
 NEARNESS A BETTER APPROACH TO CONTINUITY AND LIMITS

P. CAMERON, J. G. HOCKING, AND S. A. NAIMPALLY

This article advocates an unusual introduction to the theory of continuity and

limits. Surely an ideal presentation would begin with a simple concept having strong
intuitive appeal. That appeal should be retained along a direct line from the crude
beginnings to complete rigor. Furthermore, a student should be able to "drop off"
at any point along that line without losing all his feeling for the topic. A natural uni-
fication of related ideas should come easily as should the extensions to more ad-
vanced material. And any presentation should lend itself to class participation. We
think our approach comes closer to such an ideal than any other we have seen.
In discussing continuity the calculus teacher frequently sets a subtle trap. There
usually comes the moment when the teacher considers some fixed point c in the
domain and says that points near c are to be carried to images near f(c). Then by
looking at such a function as I /x with c = 1/1000, he may refine this by saying that
points as near c as you like must be considered. He may speak of the function having
no jumps, etc. All of this discussion has led the unsuspecting students to a "co-
variant" view of continuity. Then the teacher switches direction, hitting the students
with the "contravariant" c-6 definition. The resulting confusion leaves the class
foundering.
Another major difficulty with the standard approach is what could be called its
structural problems. From the usual c-3 definition it is not at all clear what really
lies behind the basic concepts of limits and continuity. Students are apt to think
that ordering, absolute values, etc., are involved with continuity. This makes the
transition from Calculus to Analysis very difficult-a matter which was the subject
of an MAA Panel Discussion (Winter Meeting, Denver, 1965).
Our approach to continuity and limits attacks the difficulties just outlined. The
basic tool is the topological concept of "nearness" introduced by F. Riesz in 1908.
This idea appeals strongly to the students' geometric intuition. It is very easily
introduced, a very large majority of our students grasped its essence immediately
and the passage to complete rigor is direct and readily accessible. The end result
is equivalent to continuity in terms of the closure axioms but the entire theory
remains closer to its intuitive foundations. In our experience students have more
confidence in their own understanding even when that understanding is incomplete.
We have found it considerably easier to lead a class to formulate definitions and to
contribute to the proofs. The theory extends almost trivially to arbitrary topological
spaces when necessary, and identical proofs hold in every setting. Finally, there is
a striking unification of the theories of continuous functions, of various limits, uni-
form convergence, uniform continuity, etc. Thus the bridge to intermediate level
courses is very easy to cross.

1. Nearness. Suppose 0 # A c X c R4 (or R.".) and x e X. The two notions

739

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
740 P. CAMERON, J. G. HOCKING AND S. A. NAIMPALLY [September

"x is near A" and its denial "x is far from A" are first suggested by means of dis-
cussion of special examples such as

(1) A = {!:neN) with x = 0,2, and N the set of natural numbers;

(2) A = (0, 1) with x = -1, 0, 1, 2.

Our goal here is to lead the class to a formulation of a precise definition.

1.1 DEFINITION OF NEARNESS. If A c: X c: Rl (or R') and x e X, then x is near

A in X provided that for each r > 0 there exists a e A with x - a I < r. We write
x 3 A. The denial x is far from A means either A = 0 or for some r > 0 we have

Ix-aI > r for every aEA. The symbol is x4A.

A class usually comes up with the basic properties (a), (b), (c) and (e) below
but rarely will find (d).

1.2 THEOREM.
(a) xbA => A # 0
(b) xeA => xbA
(c) xb(A uB) .xxbA or x bB
(d) xbA and VaeA, abB xbB
(e) x3{y} if and only if x y.
1.3 THEOREM. If A is finite, then xbA iff xeA.

1.4 THEOREM. x 3 A, x 4 A => x (A-F) fbr any finite set F.

1.5 DEFINITION. x c: R (or R1' ) is discrete ifl x 6(X- {x}) Vx e X.

1.6 THEOREM. Any subset of the set of integers is discrete.

We foreshadow important ideas by pointing out that the proofs of these results
depend only on the five "axioms" in Theorem 1.2. Then we provide (not necessa-
rily metric) examples of nearness satisfying the axioms. These include
1. x 3 A x e A (the discrete topology).
2. x 3 A A + 0 (the indiscrete topology).
3. The nearness on N * = 1 U {oo} from Definition 3.2 below.
4. Some example involving family membership. (Mothers-in-law always seem
to be far away by such definitions.)

2. Continuous functions. A function is continuous if it preserves the nearness

relation. This "covariant" definition is not only appealing to the intuition, it is
precise.

2.1 DEFINITION. Let c E X cD R (or R ) and f: X -+ R (or R n) be a function. Then

f is continuous at c provided that VA c X, c 3 A => f(c) bf(A). We say that f is
continuous on X if f is continuous at every point of X and if f fails to be continuous
at c, we say that f is discontinuous at c.

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
1974] NEARNESS - A BETTER APPROACH TO CONTINUITY AND LIMITS 741

A little care in the discussion will lead a class to formulate this entire definition.
Incidentally, the discussion always includes proofs of the first two statements in the
next result.

2.2 THEOREM.
(a) Constant functions ar-e continuous.
(b) Identity functions are continuous.
(c) Restrictions of a continuous functioni are continuous.
(d) A composition of continuous ftunctions is continuous.

To illustrate the methods used to prove that the continuous functions on X c: R'
to Rl form a ring, we include a proof of closure under addition. Note that we have
written the (contrapositive) proof in very terse form. This is not the way we present
it to a class. In fact, preparatory discussion results in the class constructing much
of the proof. (We have experienced an unusually high degree of class participation
throughout this entire presentation.)

2.3 THEOREM. Let ceX c: R (or R'). If both f: X -- R8 and g: X -? R are

continuous at c, so is f + g.

Proof. Suppose to the contrary that f + g were not continuous at c. Then there
would exist A c= X such that c3A but (f+ g)(c) (f+ g)(A). In other words,

for some r>O, |(f+g)(c)-(f+g)(a)| > r for each aeA, or

(*) f (c)-f (a) + g(c)-g(a) ? r for each a e A.
Define sets
A1 = {a E A: If(c) -f(a) < r/2}

A2 {asA: If(c)-f(a)I > r/2} = A-A1.

If a e A1, then we must have I g(c) - g(a) ? > r/2 or else (*) would
Thus g(c) g(A1). Since g is continuous at c we must have c A1. By identical
reasoning, we obtain c A2. Theorem 1.2(c) now tells us that c (A1 UA2). Be-
cause A1 UA2 = A, this contradicts the original assumption and completes the
proof.
We are careful to point out that such proofs depend only on the "axioms" in
Theorem 1.2. This is in preparation for our treatment of limits to be given in Sec-
tion 3. The final step in this presentation of continuity is to establish the usual e-3
characterization and discuss its utility.

3. Limits. We have not used limits, in the usual sense, to define continuity. In
order to develop the needed theories, however, we employ our knowledge of con-
tinuity. By defining a limit to be an "expected" value, the value causing a certain
new function to be continuous, all of our theorems on continuity apply directly
to provide the properties of limits.

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
742 P. CAMERON, J. G. HOCKING AND S. A. NAIMPALLY [September

3.1 DEFINITION OF LIMIT OF A FUNCTION. Let X c R (or R1P) and l

Given f: X -+ R (or Rfl) we write

lim f(x) = Ie D1 (or [Rtn),

x-+c

provided that the function f* on X u {c} defined by

f *(x) = f(x) if c G X - {c}, f *(c) = 1,

is continuous at c.
(At last we see the limx,cf(x) = f(c) property of continuous functions.)
To study convergence of sequences we append an element oo to %N and postulate
that oo is near A c: N if A is infinite. More precisely, on NdJ = NJ U {oo} we provide
a nearness 3 as follows:

3.2 DEFINITION OF NEARNESS ON N*. Let x E- N* and A c R*. Then x 3 A pro-

vided that either
(i) xe R- and xeA, or
(ii) x = oo and oo eA or A is infinite.
The students easily verify that this nearness satisfies the "axioms."
If 0: %N -? R4 (or Rn) is a sequence, we again define its limit as an "expected
value.

3.3 DEFINITION OF LIMIT OF A SEQUENCE. Let 4: NJ -e Rl (or R") be a sequence

and -let I e R (or Rl). Then-e I or

lim 0(n)= 1

provided that the function 0*: N * R l (or R n) defined by

4)*(n) = 4)(n) if n e N, )*(oo) = -

is continuous at oo.

3.4 THEOREM. Let 0 + A c X cz R (or1 Rf) and let x e X. Then x3A iff ther e
exists 4: N -+ A with 0 -+ x.
3.5 COROLLARY. Let X c: R (or R') and consider f: X -? R (or Rfn). f is con-
tinuous at ceX iff,for each 4): N -+X, c) - c =-f o c) -of(c).

In the same way, infinite limits, lim,.00f(x) are defined by producing a nearness
on R* = R u {_- o, oo} and then using obvious analogs of Definition 3.3. The
same proofs hold in far more sophisticated settings, too. Let X be a set and denote
by B(X) the space of bounded functions f: X -+ R with the metric

d(f, g) = sup I f(x)-g(x)|.

This provides a nearness bd on B(X) and if Definition 3.3 is used in this setting we
have uniform convergence!

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
1974) NEARNESS - A BETTER APPROACH TO CONTINUITY AND LIMITS 743

3.6 THEOREM. Let 4: N -* B(X) be a sequence of bounded real-valued func-

tions on X and let g e B(X). Then 4 -+ g iff for each s > 0 there exists mO e N
such that Vn > mO Vx E X
I )n(x)-g(x) I < .
The point of this program is that the usual theorems about all of these variously
defined limits are now no more than special cases of the corresponding results con-
cerning continuous functions. We need leave nothing unproven because it is too
difficult at an early stage nor need any proof be "left as an exercise" because its
inclusion is too tedious.

4. Proximities. The program sketched above does not include uniform conti-
nuity and of course it need not. Even in an honors calculus course one usually evokes
uniform continuity only to prove that a continuous function is integrable. That
usage is unnecessary (see our offer of sets of notes later.) However, in an intermediate
analysis course uniform continuity becomes important and an easily motivated
generalization of nearness applies.
It is a natural step to pass from "nearness of a point to a set" to "nearness of
a set to a set. " The latter notion, called a proximity, also was introduced by F. Riesz
in his 1908 paper.

4.1 DEFINITION. Let A,B c X cD R (or Rm). We say that A is near B in X, and
write A c B, provided that for each r > 0 there exist points a E A and b E B such
that la-b j < r. (Surely if B = {x}, this is precisely the definition of x5A).
By calling for properties analogous to those found in Theorem 1.2 the instructor
can lead the class to formulate most of the next result.

4.2 THEOREM.
(a) A5B A $A 0,B # 0.
(b) AbB => BbA.
(c) AfnB#0 =AbB.
(d) Ab(B UC) A6BorA65C.
(e) AbB and bbCCVbeBA=>AbC.
(f) If 3xeX with xbA and xcB, then AbB.

The converse of 4.2 (f) is not true. The standard counterexample is A - N and
B = {n-1/n:neN}.

4.3 THEOREM. lf A and B are finite, then A 6 B iff A n B #& 0.

4.4 THEOREM. AbB, A nB = 0 > Ab(B-F) for any finite set F.

Examples of proximities satisfying the "axioms" (a)-(f) of Theorem 4.2 can

easily be found.
1. A c1 B iff A n B # 0 (the discrete proximity).
2. A 62 B if A = 0 = B (the indiscrete proximity).

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
744 P. CAMERON, J. G. HOCKING AND S. A. NAIMPALLY [September

3. A53B iff 3x with x(5A and x(5B.

4. A (4 B if either A 63 B or both A and B are infinite.
5. On R, A 65 B if either A 63 B or both A and B are unbounded above or
both are unbounded below.
One of the motivations for Definition 4.1 is this: Starting with a topology, as
given by a metric in our case, a proximity was defined so that whenever one of the
sets involved is a singleton set, the definition reduces to that of the original nearness.
One says then the proximity is compatible with the nearness. For instance, the prox-
imities 3, 4 and 5 above are compatible with the usual nearness 1.1 on DR. In fact,
there are 2C different proximities compatible with 1.1 on 11. One sees that it is very
significant to pass from nearness to proximity.

4.5 DEFINITION. Let X c (- (or Rfm) and consider a function f: X -+ R (or R'f).
We say f is proximally or uniformly continuous on X provided that for all sets
A,B cX,
A(5B =>-f(A) (f(B).

It is obvious that a uniformly continuous function is continuous. However, the

converse is not true. For a counterexample, consider the function f: R -+ DR defined
byf(x) = x2. Let A = N and B {n -(1ln): n E NJ}. Then A ( B butf(A) b f(B)
because the distance j f(a) -f(b) is never less than 1.

4.6 THEOREM.
(a) Constant functions are uniformly continuous.
(b) Identity functions are uniformly continuous.
(c) A restriction of a uniformly continuous function is uniformly continuous.
(d) A composition of uniformly continuous functions is uniformly continuous.

The algebra of uniformly continuous functions is not so well-structured as that

of continuous functions. For instance, note that f(x) = x2 is a product of identity
functions but is not uniformly continuous on R1.
Of course, both the nearness and the e-( definitions of uniform continuity should
be known and used. However, the equivalence of the two is not a trivial theorem.
It was proved first by Efremovic in 1952 (Geometry of Proximity, Math. Sbornik
Vol. 31, pp. 189-200). Cleveland (this MONTHLY, 80 (1973) 64-66) has provided
a readily accessible proof.
We conclude this article with an offer. A reader interested in the details of the
material sketchily outlined here may obtain them from the author at his Thunder
Bay address. There are thiee sets of notes available at cost. The first covers material
on continuity and limits needed in an elementary calctulus course. A second more
extensive set of notes augments an introductory analysis or advanced calculus
course. The third set of notes on elementary integration theory was mentioned earlier.
The level of difficulty is suitable for an honors calculus course, for instance, and
the approach provides a proof of the integrability of continuous functions without

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms
 1974] THE PROFESSOR'S SONG 745

using uniform continuity. In fact, this theorem comes out naturally in a program
leading to the fundamental theorem of calculus.

The author S. A. Naimpally was partially supported by an operating grant from N. R. C. (Canada)
and a research grant from the President of Lakehead University.

DEPARTMENT OF MATHEMATICS, LAKEHEAD UNIVERSITY, THUNDER BAY, ONTARIO, P7B 5E1, CANADA.
DEPARTMENT OF MATHEMATICS, MICHIGAN STATE UNIVERSITY, EAST LANSING, MICHIGAN, 48823,
U. S. A.
DEPARTMENT OF MATHEMATICS, LAKEHEAD UNIVERSITY, THUNDER BAY, ONTARIO, P7B 5E1, CANADA.

THE PROFESSOR'S SONG

Words by Tom Lehrer - Tune: "If You Give Me Your Attention"

from Princess Ida (Gilbert and Sullivan)

If you give me your attention, I will tell you what I am.

I'm a brilliant math'matician - also something of a ham.
I have tried for numerous degrees, in fact I've one of each;
Of course that makes me eminently qualified to teach.
I understand the subject matter thoroughly, it's true,
And I can't see why it isn't all as obvious to you.
Each lecture is a masterpiece, meticulously planned,
Yet everybody tells me that I'm hard to understand,
And I can't think why.

My diagrams are models of true art, you must agree,

And my handwriting is famous for its legibility.
Take a word like "minimum" (to choose a random word), (*)
For anyone to say he cannot read that, is absurd.
The anecdotes I tell get more amusing every year,
Though frankly, what they go to prove is sometimes less than clear,
And all my explanations are quite lucid, I am sure,
Yet everybody tells me that my lectures are obscure,
And I can't think why.

Consider, for example, just the force of gravity:

It's inversely proportional to something - let me see -
It's r3 - no, r2- no, it's just r, I'll bet-
The sign in front is plus - or is it minus, I forget -
Well, anyway, there is a force, of that there is no doubt.
All these formulas are trivial if you only think them out.
Yet students tell me, "I have memorized the whole year through
Ev'rything you've told us, but the problems I can't do."
And I can't think why!

*) This was performed at a blackboard, and the professor wrote: A/\\A/\V\fVV\/\AVW

This content downloaded from 128.210.106.77 on Wed, 25 Sep 2019 15:09:00 UTC
All use subject to https://about.jstor.org/terms