Another Characterization of Inner Product Spaces

Author(s): W. A. Kirk and M. F. Smiley

Source: The American Mathematical Monthly, Vol. 71, No. 8 (Oct., 1964), pp. 890-891
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2312400 .
Accessed: 01/02/2015 00:15

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

This content downloaded from 128.235.251.160 on Sun, 1 Feb 2015 00:15:57 AM

All use subject to JSTOR Terms and Conditions
890 MATHEMATICAL NOTES [October

ANOTHER CHARACTERIZATION OF INNER PRODUCT SPACES

W. A. KIRK AND M. F. SMILEY, Universityof California,Riverside

In a recentnumberof this MONTHLY Dunkl and Williams [1 ] prove that if

x and y are vectors in a normedlinear space, then
(1) IIYIIITx- yll - (Ilxsl + SIYIy)II
411xll IIxIIy- IIyIIx11;
while if x and y are vectors of an innerproductspace, then
(2) liX- Yll>? (iIXii
2ixiiIIYII + IIyII)II - IyJxJII1.
lixily
Dunkl and Williams raise the question whether(2) characterizesinnerprod-
uct spaces. E. R. Lorch [2 ] gave the followingconditionon a normedlinearspace
X :

= lIyII
(L) If x, yGX and l|xii thenlx+a-1x+yii ?IIx+yIIforall positivereal
numbersa.
Lorch showed that (L) holds if X is an inner product space. Replacing x
and y by ax, -a'ly, respectivelyand using |lxii= lylfl one easily verifiesthat (2)
implies (L). Since Dunkl and Williams proved (2) in an inner product space,
it is clear that (2) holds in a normed linear space X iffX is an inner product
space.
The question of equality in (1), (2) was also raised by Dunkl and Williams.
A slight rearrangementof their proofshows that equality holds in (1) iffx = 0,
y=O, or x=y. For
(3) 1iflylix- 11ilyll Ilyll{iix - Yf+I I IIYII I }X
- ffxjI
(4) I IIYII- IIXIII < IIX - YIIX
(5) 11llylix- llxllyll< lisl{llix - Y11+ I IIYII- IIXlI
I }.
We easily see that equality holds in (1) iffequality holds in (3), (4), and (5);
henceiff|ix-Yi = iIIYII
-||x| I andililylix-lixiyl =1 yf0=f31xilwith 3=Jfx-yjj
+1 iyIIY
-lixilI; whichyields,fornonzerox,y, lxi| lyI and x=y. The converse
is clear.
Finally, we note that equality holds in (2) iffx=0, y=0, l|xi I=IIyii, or
xiIyiI+yIixIi = 0. This may be proved by the usual squaring process (cf. [31).
From (3), (4), and (5) one may obtain slightlymore than (1), i.e.,

(6) | ||y|ix- iixi|y|| iIYII)? 2iix11iIYilix- Y|I,

max(iixii,
an inequalitydue to J. L. Massera and J. J. Schaffer[Annals of Math. 67 (1958)
p. 538]. One cannot replace the 2 of (6) by 1 even in an innerproduct space. In
a letterto the editor,ProfessorSchafferobserves that in an innerproduct space
it is possible to replace the 2 of (6) by anythinggreaterthan 1 foriix-Yiisuffi-
cientlysmall compared to |xi|I IyIf,and that this propertycharacterizesinner
product spaces of dimensionat least three. ProfessorSchafferplans to publish

This content downloaded from 128.235.251.160 on Sun, 1 Feb 2015 00:15:57 AM

All use subject to JSTOR Terms and Conditions
1964] MATHEMATICAL NOTES 891

this resultin anotherjournal. Conditionsforequality in (6) seem difficultto ob-

tain.
ResearchdonewhileM. F. Smileyheld NSF GrantGP1447.
References
1. C. F. Dunkland K. S. Williams,
A simplenorminequality,
thisMONTHLY, 71 (1964) 53-54.
2. E. R. Lorch,On certainimplicationswhichcharacterize
Hilbertspace,Ann.of Math.,49
(1948) 523-532.
3. M. F. Smiley,The proofofthetriangleinequality,
thisMONTHLY, 70 (1963) 546.
EditorialNote.The questionsraisedby Dunkland Williamswerealso answeredin lettersto
Prof.Williamsby Gary H. Meisters,Universityof Colorado,Donald A. Sarason,Institutefor
AdvancedStudy,and by JamesP. Crawford, LafayetteCollege.

PERIODS OF MEASURABLE FUNCTIONS AND

THE STONE-6ECH COMPACTIFICATION
Z. SEMADENI, Poznafi,Poland

1. Periods of measurable functions.A number T is called a period of a

functionf definedfor - oo <t < oo iff(t+ T) =f(t) forall t. Obviously, the set of
all periods off is an additive subgroup of the reals and forany periodicfunction
we have two possibilities:Either thereexists the smallest positive period To and
all periods are of the formnTo with n = 0, ? 1, ? 2, * - * or the set of periods
is dense. We have two typical examples of functionswith dense set of periods:
the characteristicfunctionof the set of rationals and the functionf(x) -(x)
-4(1)x, where $ is any nonmeasurableHamel solution of the equation q(x+y)
=4(x) +4(y); in both cases any rational numberis a period. Obviously, a con-
tinuous functionwith dense set of periods must be constant. There exist a
nonconstant measurable functionwith an uncountable set of periods and a
measurable functionwith Darboux propertyand a countable dense set of peri-
ods (cf. [4] pp. 833-836).
The purposeofthis note is to give a simpleproofofthe followingtheoremdue
to A. Lomnicki ([4], Theorem 5).
THEOREM1. If a measurablefunctionhas a denseset ofperiods,thenit is con-
stantalmosteverywhere.
The proofis founded on the followingwell-known
LEMMA. If T is a positive function
periodofa measurable f andfo'If(t) dt< oo, f
thenthelimit
1 f2
9(f)-= lim_
X 2x
f(t)dt

existsand is equal to T-1fTf(t)dt.

This content downloaded from 128.235.251.160 on Sun, 1 Feb 2015 00:15:57 AM

All use subject to JSTOR Terms and Conditions