Beyond Sin and Cos

Author(s): Martin E. Muldoon and Abraham A. Ungar

Source: Mathematics Magazine, Vol. 69, No. 1 (Feb., 1996), pp. 3-14
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2691389
Accessed: 09-09-2015 18:14 UTC

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 [email protected].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to Mathematics
Magazine.

http://www.jstor.org

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
 ARTICLES
BeyondSinand Cos
MARTIN E. MULDOON'
YorkUniversity
NorthYork,OntarioM3J1P3, Canada

ABRAHAM A. UNGAR
NorthDakota StateUniversity
Fargo,ND 58105

1. Introduction
The higher-order circularand hyperbolic functions
deserveto be betterknown.Here
we givetheirmainpropertiesin orderto makethemmoreaccessibleto teachersand
studentsin calculus,linearalgebraand differential equationscourses.The studyof
thesefunctions can be relatedto such diversetopicsas thebinomialtheoremand the
fastFouriertransform.
Here,foreach positiveintegern, we definer functions FJ,(x), r = 0,1
.I., n-1.
The cases a = 1 and a = -1 correspond,respectively, to whatare usuallyknownas
generalizedhyperbolicfunctions and generalizedcircularor trigonometric functions.
We findit usefulto retainthe parametera; the case a = 0 also gives something
interesting.
The functions consideredhereare elemnentary and can be a richsourceforstudent
projectsand investigations.

2. Background
The trigonometric functionscan be generalizedin manyways,some of themindis-
pensable in the applicationsof mathematics. We mention,forexample,the Bessel
functions[40], specialcases of which have been aroundsincethe timeof Euler, the
hypergeometric functions[30] and theirvariousgeneralizations.
But the deeperstudy
of thesefunctions becomesdifficult veryquicklyand so theyare littlestudiedexcept
by thosewho need themforsome application.It is of interesttherefore to note that
there existsa class of functionsthat preservethe elegance and simplicity of the
trigonometricfunctions and thatis easilypresentedto studentsin elementary courses.
These generalizedcircularand hyperbolicfunctionshave been rediscoveredoften
sincethefirstrecordedaccountbyVincenzoRiccatiin 1757. Theypreservethe flavor
of striking
resultslike Euler's formula

eix = cos x + i sinx, (1)

'The wvork ofthisauthorw'assupportedbya grantfromthe NaturalSciencesand Enginieerinig

Research
Counicil(Canada).
3

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
4 MATHEMATICS MAGAZINE

the determinantal
identity
Cosx sinx=1 (2)
-sin x cos x
and the matrixidentity

(Cos
x sinx (Cos y siny _(cos( x+ y) sin(x+ y)
-sin x cos x -sin y cos y -sin(x + y) cos(x + y) (3)

They also satisfydifferentialequations similar to the equations y' - y = 0,

y" - y = 0, and y" + y = 0, satisfiedby the exponential, and trigonomet-
hyperbolic,
ricfunctions, respectively.
In spite of theirsimplicity and the lengthof time theyhave been around,the
generalizedcircularfunctions are seldomdiscussedin textbooks.(An exceptionis [141,
pp. 336-339). This maybe because theydo not seem to have obviousapplications.
(See, however,the recentreferences[28],[42],[20],[15],[9].) We believe that the
value of the studyof these functions lies ratherin theircompellingintrinsic
beauty
and in providinga rich source of examplesand motivationin variouselementary
courses.

3. The GeneralizedHyperbolic
Functions
We can definethe exponential
function
by the usual sum
x0k 2 x3
exp(x)= E 7=1+X+ 2! + ! +...
k-o

Inthecaseofthehyperbolic
functions
wetakeevery
second inthissum:
term
a) 2k+1 3
sinh(x) = E 2 =
A I

r
cosh(x)= E x2k
X =
x2
E
k=O (2k)!

Anobvious waytogeneralize
thisistotakeevery
nthterminthesum.Thus,forany
positive
integern,we define
then a-hyperbolic
functions
ofordern, F,l,,onthe
realline(orinthecomplex
plane)bytheinfinite
powerseries,
o ak
Fnr(X)E ( nk+r) x!k+

xr axn+r a2x 2n+r

r! (n+r)! + (2n+r)! +
r=0,1,...,n-1,
wherea is real (or complex).For consistency
we take

F,0(0) = 1

The functionFar is called the a-hyperbolic

function of ordern and rthkind.
There is a single a-hyperbolicfunctionof order 1; it is the exponentialfunction
F1ao(x)= eax. Thereare two a-hyperbolic of order2; theseare the
functions

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
VOL. 69, NO. 1, FEBRUARY 1996 5
functionsF20(x) = cosh(va x) and F2a1(x)= (1/a)sinh( ax), where a is an
specifiedsquare rootof a, givingrise to the circularfunctions
arbitrarily (a = - 1)
and the hyperbolicfunctions(a = 1) as special cases. Similarly,there are three
a-hyperbolicfunctions of orderthree,and so on.
Writingthe infiniteseriesin the definition
of F,", explicitly,
one mayreadilyfind
the relationship

1,,r(X) = Fa 4) (tO 4

n
where a is an arbitrarily specifiednthrootof a. Although(4) showsthatthe case
a:# 0 can be reducedto the case a = 1, retaining the a preservesa certainelegance
in theformulas.It is also interestingto see, in Section5, howthecase a = 0 givesrise
to polynomials;withoutthe explicituse of theparametera, thisconnectionmightgo
undetected.Like thecircularand hyperbolic functions,the a-hyperbolicfunctionsare
generatedbyparticularly simpledifferential equationsnormalizedby "natural"initial
conditions.The a-hyperbolicfunctionof ordern and rth kind Flr( x) satisfiesthe
differential
equation

f(1)(x-) = af(x) (5)

normalizedby the initialconditions

f(k)(o) J, k # r, 0<k ?n -1,

\1, k =r.
Moreover,differentiation
permutesthe a-hyperbolic
functions
cyclically,
apartfroma
factora in one case:

d fFrO(x) < r< n-1,

Tj1j r(X) a ,
\F, 1(x) r=0.

4. The GeneralizedEulerFormula
Our definition
can be shownto lead to the generalizedEuler formula

e lx = n-
E (4 1,,r(x)(6
r-O (6)
n
wher
where 4 is an arbitrarily specifiednth rootof a. Obviously,thisreducesto (1) in
case n = 2, a = - 1. Since thereare n nth rootsof a, we see that(6) is actuallya
systemof n linearequations.In Section9, we willuse theFouriermatrixto showthat
the system(6) can be solvedforthe F car r= 0,..., n-1, to give

F,~ ~ n ~ ~ k-orn
-rrk
= crkxp[w kVx] (7)

Here w,,= exp[2iri/nIis a primitiventhrootof unity.Actually, in thecases a = ? 1,

(7) has been used to dfine the a-hyperbolicfunctions;see, e.g., [13], p. 212.
Formula(7) may be used to expressall the a-hyperbolicfunctionsin termsof
trigonometric and exponentialfunctions.In the case a = 1, the three a-hyperbolic

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
6 MATHEMATICS MAGAZINE

of order3 are
functions

0( x) =
F3" [e x+ 2e x/2 cosa2j]

F3,1(x)= .[ex-2ex/2cos( 2 + 3

(
and

F31,2(x)= . [eX-2e CX/2cos - 32)]

The firstof thesefollowseasilyfrom(7). The otherscan be obtainedthenfrom(7) or

of the firstone. FIGURE 1, producedusingGNUPLOT2,
by successivedifferentiation
givesgraphsof the hyperbolic functionsF3 k(X) fork = 0,1, 2. In the case n = 4, we

200
150-

100-

50

-50
-12 -10-8 -6-4 -2 0 2 4 6
FIGURE 1
Fd3k(X)vs. x,k=0,1,2.

get the elegantformulas

Flo (x) = (1/2)(cosh x + cos x),
F441(x) = (1/2)(sinh x + sinx),
= -
F41,2(x) (1/2)(cosh x cos x), (8)
F4,3(x) = (1/2)(sinh x - sin x).

5. The a-hyperbolic
Matrix
To exhibitthe generalization
of identities(1) and (2) providedby the a-hyperbolic
we definethe n X n a-hyperbolic
functions matrixH,O(x) by the equation
F
l)(x) x) ...
Fl0(x) ((x) ... Fa 1-2 X
caF)- FO% (x) F",1(x) F),11(x)
H)c ( Xa) F 2(X) a . a )I-3(X) (9)

aFna (x) aF%2(x) aF,3(x) ... F

The a-hyperbolicmatrixH,"(x) is factor-circulant,thatis, it is a matrixobtainedby
by a each of the elementsbelow the maindiagonalin a circulantmatrix.
multiplying
Factor-circulants
are consideredby Ruiz Claeyssenand dos SantosLeal in [28]; see

(C) 1986, 1987, 1990, 1991 Colin Kelley,ThomasNVilliams.

2Copyright

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
VOL. 69, NO. 1, FEBRUARY 1996 7

matrixgeneralizesthenotionofcirculantmatrix
also [27]. The factor-circulant (a = 1)
and skew-circulant matrix(a = - 1) [10]. Circulantmatriceshave interesting and
usefulproperties([101,[6]) thatare sharedby factor-circulant
matrices,e.g., foran
arbitrary matricesforma ringundermatrix
fixedn and a thefactor-circulant addition
and multiplication.
In [36], the name "a-hyperbolicmatrix"was applied to the transposeof H,,(x),
ratherthan to H,,(x) itself.The resultingmatrix,the resultof multiplying the
elementsabove the maindiagonalin a circulantmatrixby a was describedin [36] as
factor-circulant.Our currentterminology seems more appropriatein view of the
notationused in [28] and the notationforskew-circulantmatricesin [10]. It follows
fromwhat was shown in [36], that the a-hyperbolicmatrixH,,(x) satisfiesthe
identities
IH,(X)l = i, n>1 (10)
and
Ha(X) H,O y)=H,x +Y), n>1, (11)
forall real (or complex)x and y, where IH,O(x)lis the determinantof the matrix
H,( (x), and where denotes matrixmultiplication. We will see shortlyhow these
resultsfollowfroma studyof differential
equationsin matrixform.Identities(10) and
(11) reduce to identities(2) and (3) when n = 2 and a -1. When n = 1, (11)
reducesto the exponentialidentity
eoixeol = e0l"Y).

6. A Connection
to theBinomialTheorem
For n > 0 and a = 0 (withthe usualconvention
that0? = 1), the a-hyperbolic
matrix
H,W(X) is uppertriangular,
X2 ()1-l 1 -
1 x .. (n- 2)!

H,?(x) = |0
H)10(x) 1 X(n.-2)! |(2 (12)
~~~(n-3)
0 0 1 (n-3)!

0 0 0 ... 1
fromwhichidentity (10) is obvious,and identity
(11) is a matrixformof thebinomial
theorem;see [36] and [17]. This can be seen by considering

x 2! ... ]y 2 ...
0 1 x ..O 1 y ...
O O 1 .. O 1 ...
. . *. ... ... ... .. ... ...
2
x+ (+Y

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
8 MATHEMATICS MAGAZINE

Everyentryabove the diagonalin the right-hand matrixis of the form(x + y)k/k!,

and it is obtainedby multiplyinga rowof the firstmatrixon the leftby a columnof
the secondmatrixon the left.This productis seen to containexactlythosetermsthat
arisein the binomialexpansionof (x + y)k/k!.Thus we havewhatamountsto a link
betweenthe binomialtheoremand the theoryof ordinary differential
equations;this
has been emphasizedin [34] and [37]. The binomaltheoremhas a longhistory [7] and
is foundthroughout the mathematical literaturemainlyin treatiseson combinatorial
and numbertheory.In literature
analysis,statistics, on differential
equations,however,
it usuallyappearsonlyas a tool;see, e.g., [31], p. 404.

7. Differential
Equationsin MatrixForm
H,' is the unique solutionof the matrixdifferential
equation
0 1 0 ... O
O 0 1 ... O
M'=[: : : . :M
O O O ... 1

M(O)= I. Thatitis a solution

thatsatisfies is easytoverify.
Itsuniqueness
follows
from standard
resultsinthetheory ofsystems ofdifferential
equations;
see,e.g.,[5],
Ch.3. In fact,
ifwewrite
0 1 0 ... O
O 0 1 ... O
A=[. 0. ... I
O O 1
La ...

wefindthat
H,'(x) =exp( Ax),
withtheusualnotation fortheexponential
ofa matrix
[5,?3.11].Bytheusualtheory,
theWronskian ofsucha matrixisconstant
sowefindthatdetH,"'(x) = detH,l'(0)= 1,
i.e.,theequation(10).Wemayalsorecover (11)with
thisformulation:Foreachfixed
y,bothsidesof(I1) arematrix solutions
ofthesystem M' = AM,satisfying M(0)=
F,l(y). Hence,byuniqueness, theyareidentical.
It is ofinterest
alsotoremark that
A"l= aI. (13)
Onewaytoseethisis bydirectcomputation.(Taking successivepowersof A causes
an upwardshiftin boththe I's and the a's.) It is also a consequence
of the
Cayley-Hamilton
theorem,accordingto whichA satisfies its owncharacteristic
equationdet(AI-A)= 0; expanding by the firstcolumn, det(AI-A)= Ak+
( _ 1Y,- 1 (-C a)(- 1)' - 1 = k' -C o.
Equation
(13)shows
thatH,asatisfies
then-th
order
matrix
differential
equation
M()- = aM
analogous
tothenth-order
scalardifferential
equation
(5) satisfied
byF,ar(x).In fact

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
VOL. 69, NO. 1, FEBRUARY 1996 9
H,a is the unique matrixsolutionof the system

M(")= aM, M(k)(O) =A', k=0,1..,n-1.

Anotherway to expressthe relationship between H,,(x) and A, motivatedby the
specialcases in [14], pp. 337-339, is
n-i

Ha( (x) = E F,'k(x)Ak.

k-o

We remarkthatin the case a= 0, the matrixH,?(x) givenby (12), has the familiar
form[5] of the matrixeAx, when

0 1 0 ... O
O 0 1 ... O
A= : : K. :
O O O ... 1
LO O O ... O

8. MiscellaneousResults
Thereare simpleformulas fortheLaplace transforms of thegeneralizeda-hyperbolic
functions. The Laplace transformof F,,,r(at), 0 < r < n, n = 1,2,. .., wherea and a
are real constants,
is
00 r
i~~~~i-r-Ia
fe -stF,a,r(at) dt = s" a(14)
o ~~~s
+ aan~(4

Perhapsthe simplestway to see thisis to use the differential

equations(5) and the
well-known relationsforLaplace transformsof derivativesof functions.
It shouldbe
mentionedthatthe generalized1-hyperbolic functionsare relatedto the Mittag-Lef-
flerfunction
oo k
x) = E F(ykx + 1)'
E,(E()=k-o

whichis important
in the theoryof entirefunctions.
The relationis F,,
'O(x) = E,,(x").

9. The Fourier
Matrix
The FouriermatrixJg is definedby

1 ... 1
1 (t)1 Cl)
2
. (X9n-1?-
1= 11 @2 (Xp4 Co2(13-1) (15)
... ...111 ... 3 ... ...
13
- 2
1
11-)ll (,2(n 1) ... C
11- 1

where co,=2 nrI/n is a primitiventh rootof unity.This is the matrixthatarisesin

connectionwiththe discreteFouriertransform; see, e.g.,[32] and [38]. Y, connectsa

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
10 MATHEMATICSMAGAZINE
columnvectorof complexnumbersy= [Yo, Yi> ., y11_ ]T, withits "Fouriercoeffi-
cients"c =[co, cl. c l]T bytheequations
,C

y =5';, c = ;7-' (16)

It is a fundamental fact,easilyprovedby using 1 + w,,+ w,1
+ . + W," -' = 0, that
n1= 1 , wherethe bar denotescomplexconjugate.Thereare at leasttwoways
in whichthe Fouriermatrixis connectedwiththegeneralizedhyperbolic functions.In
the firstplace, equation(6) holds,whereVa is an arbitrarily
specifiednthrootof a.
Since the nthrootsof a are of the form
-
W
XC9 a t)ttz 4

(6) is actuallya set of n equations,whichin matrixformmaybe written:

edX F,aX)
e ,,'WX =i,,: nG F,0 ( X)

e 1~l>xa4 F,aI-1( x)

Using - = n- , we easily invert(6) to get (7). Battioni[2] gives the relations(8)

betweenfunctions
of orders4 and 2, a specialcase of the moregeneralresult

F2 ,r(X)=[F,nr(X)+ F,,7',(X)I/2, r=0,1,...,in-1, (17)

F2,nr+in(X) =
[F,nr(X)-F,-',(x)]/2, r=0,1,...,in-1, (18)

givenin a differentnotationin [13], (33), p. 216. Formulas(17) and (18), which

expressgeneralizedhyperbolicfunctions of even orderin termsof thoseof halfthis
order,can be provedin severalways.One wayis directly fromthe seriesdefinitions.
Anotheris. to show that the functionson both sides of (17) and (18) satisfythe
differential
equation
f(21n)( x) =f x), (19)

whilebothsides of (17) satisfy

the initialconditions

f1(k)(0) 0, kor, O<k<n-1

'~~'\ 1, k= r

and bothsides of (18) satisfy

the conditions

f(k)(o) = O, O< k < n - 1.

It is instructive
to see how the proofof (17) and (18) can be deduced fromprecisely
the same techniqueas is used in explainingthe "fast Fouriertransform" or FFT
([32],[33]). (The aim of the FFT is to performthe multiplications in (16) quickly,
when n is large,by halvingn repeatedly. We illustrate one such step,withn = 2In.
The processcan be repeatedif n is a powerof 2.) We have from(7),
1 2m- 1
)
2., r(X)
F2m xj
J =exp[ a)2 x] (20)
k=O

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
VOL. 69, NO. 1, FEBRUARY 1996 11
or,in termsof the Fouriermatrix,

[ 2m,0( X) ex

Flm1( 1 - e W2n,X

F2m2mi(x) x J [eW X

The idea behindthe fastFouriertransform

is thatthe matrixmultiplication
in the
equation

Yo co

[
Yi C1
I 2m 31,

Y2m-1 C2m-1J
can be achievedin slightly
morethanhalfas manystepsas wouldbe needed in direct
multiplicationby splittingthe vectoron the right-hand
side into even and odd
components, multiplyingthese half-sizevectors
,
Ceve,i = [CO)C2. C2m-2] Codd [Cl,C3, C2m-1]

byY), and puttingthe resulting

vectors
=
u= [Uo, ..... ,Un-1] mceven) V= [vO. Vm_1]
=i mCodd

back togetherby taking

y1=t+j+ v1, j = 0,1. n-i,

Yj+m=tLj V
0,1.n
W2)Vj j .

In the presentsituation,
we have

ceveii =le[Ix e . e ] Codd = [ew2x, eL . e2

Then, using to24m we see that(7), with a = 1, a= 1, and a= -1,Va=

= CUM' o2m
respectively,
gives

Fm of x
X

u=mn l( X)
Mn

FU m-[ l ( X)

and
[ Fl-j3(x)
v in
Fn 0o

C02mFrnl(x)
(02rn Fm,m-1(
I
Usingthiswe get(17) and(18).

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
12 MATHEMATICS MAGAZINE

10. Graphsand Roots

Here we confineourselvesto the case a = ? 1. See FIGURE 1 forthe graphsin the

case a = 1,n = 3. Battioni[2] providesgraphsin a fewcases forsmallvaluesof x. Of
coursecertainresultsare immediate.Fromthepowerseries,when n is even,F,,;r(x)
is even or odd accordingas r is even or odd. In case a = -1, F,, is a periodic r

functionof x forn = 2 but is knownnotto be periodicforlargervaluesof n [13, p.

216]. This can be seen (also for a = 1) by usingFloquet's theorem[3, p. 325]. Of
course,it mayhappen thatsome particularcombination of higher-orderfunctions is
periodic;forexample:
Fl 0( x) - F'2,2(
F4,o(X) X ) =ox
= COS X.

The exponentialfunctionhas no roots.The circular(hyperbolic)functions have only

real (purelyimaginary) roots.What about the higher-order functions?
F,;,'(x) has a
r-foldreal rootat the origin,but the questionof how manyotherreal rootsthereare
is moredifficult.
Resultsof Mikusiinski [22] showthateach suchfunctionhas infinitely
manypositiverootsand thatthe smallestpositiveroot xl satisfies

(r+n-1)! < < 2(r+n-1)!

(r-1)! - 1 (r- 1)!

The graphsprovidedby Battioni[2] indicatethatthe lowerbound is quite sharp.A

deeper studyof the complexrootshas been undertaken by H. Alzer[1] whoprovides
two new proofsof a resultof G. P. Meyerthat F,1'r(z) has infinitely manyroots.
Generalizingresultsof G. Polya and of A. Wiman,Alzershowsthatall the rootsof
Fn,r( z ) lie on n raysstarting
at 0 and passingthroughtherootsof z" =-1. He shows
also thatthe nonzerorootsare simple.
FIGUIIE 1 wouldseem to indicaterelations suchas equalityof F O(x) and Fl l(x) at
the rootsof F, 2(x). But thisis an illusion,based on the factthatex is negligiblefor
even moderatenegativevaluesof x.

11. History
The historyof the a-hyperbolicfunctionsis a fascinating tale of discoveryand
rediscovery.Most of the historydeals withjust the cases a = ?1. The paper by
Kaufman[19] revealsthatthese functions were dealtwithby VincenzoRiccati[25],
son of the better-known Jacopo[23] as earlyas 1757. (Note thataccordingto Katz
[18] the usual trigonometric functionsdid not entercalculusuntilabout 1739 and
thenas a resultof Euler's effortsto solvelineardifferentialequations.)The general-
ized functionsalso appear in theworkof H. Wronskiin 1811 and, accordingto [19],
are the subjectof a chapterin his book [41]. Wronski'sname is a familiarone to
studentsofdifferentialequations,butonlythrough givinghisnameto the"Wronskian."
See the accountsgivenin [12], [11, pp. 57-59] as well as the remarkin [29, p. 78n]
forfurther information on thisfascinating character.The firstjournalarticleon the
generalizedfunctions and hence the firstaccountto be stillreasonablyavailable[24]
appearedin 1827 in the secondvolumeof Crelle'sjournal.Kaufmancitesmorethan
two dozen 19th-century referencesto the functionsand a similarnumberin the first
halfof the 20thcentury.Some of these references are to rediscoveries.
Thoughit is
hardto quantifysuch an opinion,it seems to us that,comparedto similarbibliogra-
phies on othertopics,a large proportionof the referencesare to obscuresources,

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
VOL. 69, NO. 1, FEBRUARY 1996 13
thoughmany(such as [39]) are easily available.One mightexpectthat,withthe
passage of time,these functionswould findtheirway into standardtextbooksand
referencebooks. For the most part this does not seem to have been the case. A
notableexceptionis the fivepages in Chapter18 in thelast[13] of thethreevolumes
arisingfromthe BatemanManuscriptProjectat theCaliforniaInstitute ofTechnology
in the late 1940s.Nevertheless,
the functionshavecontinuedto be rediscovered,e.g.,
in 1969 by Battioni[2], in 1978 by Ricci [26], and in 1982 by one of the present
authors[35]. Theywerediscussedin 1987 byKittappa[20], in 1988 byGood [15], and
in 1989 by Coonce, Strachan,and Wiest[8]. The idea of usinga generala to unify
the specialcases a = ? 1 was used by the secondauthor[36] in 1984.
It is interestingto quote some of Davis' remarkson circulantsin the Preface
to [10]:

... The theoryof circulantsis a relativelyeasy one. Practicallyevery

matrix-theoreticquestionforcirculantsmaybe resolvedin "closed form."
Thus thecirculants constitute
a nontrivial
butsimplesetofobjectsthatthe
readermayuse to practice,and ultimately deepen,a knowledgeof matrix
theory.
Writerson matrixtheoryappearto have givencirculantsshortshrift,so
thatbasic factsare rediscoveredoverand overagain....

It seems that, mntatatis these remarksapply also to the generalized

mntatandis,
hyperbolic and circularfunctions.
The first-named authorfirstcame across these functionsin reviewingBattioni's
paper [2] in 1972 (see Math. Rev. 45 (1973), #599) and was unawareat thattimeof
the longhistory of thesefunctions.He was awareofit bythetimehe reviewedRicci's
paper [26] in 1980 (see Zbl. fir Math. 423.33008). The second-namedauthor
publisheda note[35] on thesefunctions in 1982 beforesome of the references
given
hereweredrawnto hisattention bythefirst-named author.We hope thatthisaccount
will help to make these functionsmorewidelyknownand thatotherswill be led to
studythe deeper propertiesof thesefascinating functions.
Of course, apart fromthose mentionedin the Introduction, there are kindsof
generalizationsof the circularand hyperbolic functionsthatare different
fromthose
describedin thispaper. In thisconnectionwe mentionreferences[4] and [21], for
example.
Acknowledgements. WVe
are gratefulto severalrefereesforcorrectiolns
alndcolnstructive
commiients.

REFERENCES
1. H. Alzer, Die Nullstellender Hyperbelfunktionen hoeherer Ordnung, Dissertation,Rheinisch
Friedrich-NVilhelmsUniversitat,Bonn, 1983.
2. G. Battioni,Su una generalizzazionedelle funzioniiperbolichee delle funzionicircolari.Riv. Mat.
Univ.Parma(2) 10 (1969), 39-48.
3. G. Birkhoff and G.-C. Rota, OrdinanyDifferentialEquations,3rd edition,JohnWiley& Sons, Inc.,
New York,1978.
4. B. A. Bondareniko,Generalizedhyperbolic functions.Dokl. Akad. Nauk UzSSR, 1978, 10-13.
5. M. Braun,Differential Equationsand theirApplications,
3rdedition,Springer-Verlag,New York,1983.
6. R. Chalkley,Circulantmatricesand algebraicequations,thisMAGAZINE 48 (1975), 73-80.
7. J. L. Coolidge,The storyof the binomialtheorem,Amer. Math. Monthly56 (1949), 147-157.
8. H. B. Coonce, R. A. Strachanand M. NViest,
True trigonomiietry,
talkgivein
at MAAFall Meeting,North
Dakota State University,October27-28, 1989.
9. H. B. Coonce, Permutablepolynomials forseveralvariables,AequationesMath. 44 (1992), 292-303.

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
All use subject to JSTOR Terms and Conditions
14 MATHEMATICS MAGAZINE

10. P. J. Davis, CirculantMatrices,JohnWiley& Sons, Inc., New York,1979.

11. P. J. Davis and R. Hersh,The Mathematical Experience,Birkhauser, 1981.
Basel-Boston-Stuttgart,
12. J. Dobrzychi,J6zefMaria Hoene-Wronski,Dictionaryof ScientificBiography,15, pp. 225-226,
Scribner's,New York,1978.
13. A. Erd6lyi,WV. Magiius,F. Oberhettinger and F. G. Tricomi,HigherTranscendental Functions,Vol. 3,
McGraw-Hill,New York,1955.
14. A. E. Fekete, Real LinearAlgebra,MarcelDekker,Inc., New Yorkand Basel, 1985.
15. I. J. Good, A simplegeneralization of analyticfunctiontheory,Expo. Math.6 (1988), 289-311.
16. E. AtleeJackson,Perspectiveson NonlinearDynamics,Vol. 1, CambridgeUniversity Press,1990.
17. D. Kalmanand A. Ungar,Combinatorial and functionalidentitiesin one-parameter matrices.Amer.
Math. Monthly94 (1987), 21-35.
18. V. J. Katz,The calculusof the trigonometric HistoriaMath. 14 (1987), 311-324.
functions,
19. H. Kaufman,A bibliographical noteon the highersine functions. ScriptaMath.28 (1967), 29-36.
20. R. Kittappa,A generalization of the rotationmatrixand relatedresults.Lin. Alg. Appl. 92 (1987),
251-258.
21. A. I. Markushevich, The remarkablesinefunctions,trans.by ScriptaTechnica,trans.edited by L.
Ehrenpreis,AmericanElsevier,New York,1966.
22. J. G. Mikusiliski, Ann. Soc. Polon. Math. 21(1948), 46-51.
23. A. Natucci,VincenizoRiccati,Dictionanyof Scientific Biography,11, pp. 401-402, Scribner's,New
York,1975.
24. L. Olivier,Bemerkungen Ubereine ArtvoniFunctionien, welche*ihnliche Eigenschaften haben,wie die
Cosinusund Sinus,J. ReineAngew.Math. 2 (1827), 243-251.
25. V. Riccati,Opusculumsextum:Epistolaetres,in quibus aequationesaliquae differentiales evolvuntur
per series,Opusculorumn ad res Physicas,& Mathemnaticas pertinentium,Vol. 3, Bologna,1757.
26. P. E. Ricci,Le funzionipseudo-iperboliche e pseudo-trigonometriche, Publ. Instit.Mat. Appl. Fac.
Ingr. Univ.Stud. Romna, quadernono. 12 (1978), 37-49.
27. J. C. Ruiz-Claeyssen,M. Davila and T. Tsukazan,Factor block circulantand periodicsolutionsof
undampedmatrixdifferential equations,Mat. Apl. Comput.3 (1983), No. 1.
28. J. C. Ruiz-Claeyssen and L. A. dos SantosLeal, Diagonalizationand spectraldecomposition of factor
blockcirculantmatrices,Lin. Alg. Appl. 99 (1988), 41-61.
29. G. F. Simmonis,DiJferential Equations withApplicationsand HistoricalNotes, McGraw-Hill,New
York,1972.
30. L. J. Slater,Hypergeornetric Functions,CambridgeUniiversity Press,Cambridge,UK, 1960.
31. M. R. Spiegel,AppliedDifferential Equations,Prentice-Hall, Englewood,NJ,1961.
32. G. Strang,Linear Algebraand its Applications,3rd edition,HarcourtBrace Jovanovich, New York,
1988.
33. G. Strang,An Introductionto Applied Mathematics, RVellesley-Cambridge Press, Cambridge,MA,
1986.
34. A. Ungar,Additiontheoremsfor solutionsto linear homogeneousconstanit coefficient
differential
equations.AequationesMath. 26 (1983), 104-112.
35. A. Ungar,Generalizedhyperbolicfunctions, Amer. Math. Monthly89 (1982), 688-691.
36. A. Ungar,Higherorderalpha-hyperbolic functionis.
IndianJ. Pure Appl. Math. 15 (1984), 301-304.
37. A. Ungar,Additiontheoremsin ordinarydifferential equations,Amer. Math. Monthly94 (1987),
872-875.
38. J. M. Walker,FourierAnalysis,OxfordUniiversity Press,New York,1988.
39. L. E. Ward,Some functionis anialogousto trigonometricfunctions, Amer. Math. Monthly34 (1927),
301-303.
40. G. N. \Vatson,A Treatiseon theThieony of BesselFunctions,2nidedition,CambridgeUniversity Press,
Cambridge,UK, 1944.
41. H. Wronski,Reforme Absoluedu SavoirHumnain, 3 Vols., 1847-48.
42. AV.AV.Zachary,An inversescattering formalism forhigher-order differential
operators.J. Math. Anal.
Appl. 117 (1986), 449-495.

This content downloaded from 134.129.67.114 on Wed, 09 Sep 2015 18:14:32 UTC
View publication stats
All use subject to JSTOR Terms and Conditions