Vous êtes sur la page 1sur 68




I Fondati in anul 1895

\\ ---<-1.\a\-
Gazeta Matematicd" seria
A, pubrishes research articles
mathematics' e '""rio' i""oi"d-,o in both elementary and
,"i:il::f* o.oro""d problems and solutions is
readers span a broad spectrum
*",1"#l:::::1, of rnathematical interests.
The journal is ". il{;, o?
and include professional
reviewed bv Zentrarblat "i,"ii ".jr"giate
fiir Didaktiu 0". nr.rn",il'r'egiate
,n.,I"!t;i't";;*'i,';::I[u:*:ffffi1:: "'*"'i ,i"'i",* source and a pdr n,e
Matematicd" s"*t-i,-i*Vrii*n"o
so"i?tr"to quarterry by the Romanian
The annual subscription rate
is 40 EUR/50 u-s. dollars.
the beneficiarv societatea.de The deposit should be made
$tiin(e il","r".ai.a" ai, n.-arl". for
0013 5732 0003 (for
..'*o6, l..ornt RoTg RNCB 0076
comerciardL Romana. Firaia ) lyf ,i*Eil.oozo oo+j"iiiz.iool"i ?il. usD). open
sect"oli n._riri^. iri.lir,""o" RNCB at Banca
the pavment, the aetairs "r.I"u,,
o,."r'i"'rrrr".ru", rr.-I RoBu 850
.o'o**l should be sent br.

"*,ffij:, !!!::;$,;;;;[* J;31i::;.j:;:f, it:L.."",;:

Revista este aclre
in nra,ematica e,emen,ard
sioniqti, a*.;ri, ,,,aiit',t-ru:liff11ill,l,1i-;,1,".:,?:,""Ji "i -"i"-",icii: matematicieni profe-
Revista este recenzatd de
Zentrarblat ftir Didaktik
Nlaterialele proor der Nr.rthematik.
Cazeta lvlatematicd,, Seria
A apare trimestriai.
Abonamentele se realiza prin prara bancard
tttatematice'ai;';;,r"J pent-r beneficiarul Socieralea
cle sliinte

"":",ilir*'"[lil,ilf ff *it,iq]]$tti.':l'.::i:;riggt'-s"s#1ij#s
t -r;;,;;;;iainrta"
prard. ir.prerr;lT.::
t*i-rniteqi' pri, ,:*:
,.,.n poqrd sarr prin
r;i adresa la care revistr rd tereron r/*''
A b on aq ii cr h ;ffi:u;*y,',ffi T jlff:'r:L.,]o, J;'ast " "ii:'i

crpu, cir.rx popescL,. \\,L..ro^rrn

i:::1T;"t}L::"','flf:'rl^T,1"','xlru.rr Bosro,r
Redactor geft ConNal BAolrce
Editor al rubricii de probleme:
Nrcu BsLr
ro'lt At-su. AIrHrr cur;. D.r,rer
iffi;r|,,:Hi:::I: Lesxr<,. ArexexoRu Grc,r,

'""tr,.:s:lr str. fia."i"n,i"i
Redaclia Ei administralia: expecriard pe ar,'esa of
f ice@rms. unibuc. ro.
nr. 1-1. cod o-rrrli'6,,.r.eqti.
Tel: 02r Jr4.16.53. Tel./fax:
oziirz.ii.i)- "
E-mail: off ice@rms. unibuc.
lrww. rrls . unibuc . ro

ANUL XXXVI (CXV) Nr. 1 2/ 2or8


Computing exponential and trigotrotnetric functions of

matrices in .\ 1- -.
()rrr,ir F- .

Abstract. [rL this papcr \\1:' givr ,. :-' ' .' - :' -:,' r:iji rtlation o{'
tir, rrralr'ir ,,r1,,,1' ttlial frtn, ti,,r ... '. ':.:''1. '111o-1ie
flrnctions sin;1 tucl cos jl . \\'lrcr(i 1a.'. 1- - ' .' :'-'l:r:ilti lhc'real
:.alal tililllil)le ,,l tut al jull ;rlr,l , . - ':.:
|eal cilt'trlaltt tttatrix aii(l a i\'llllir,r': ll -: .

Keywords: S, 111;114 ltr r' .' - :

ttiatLir lIig()l11)lIlatli( tlltri ti,,lr- 'l-. :-'. -

NISC: lr \ lti.

. .. , . :-_r'-.-:

.-1 -. - - .--" 1-l-i(111(' fol the calt'ttla-

ratrrrell. i. 1,r.., . :- .' ....--','Hatlriltc,tr Theoretn anci

tlre yrou,er s('ri(': ri-:- -'.1-:- l- . : - I ---- ,. nllll tllc tf igonometric fl-ulc-
tiotis. The orgarliZtttiL-rl- - -- ' :l :l ' - -' 'ii'ru':: in the first section we
gir-e fornulas lirr e1. sttt -f ':, . - -- -t -'ttr-: ll the cigcrlr.tilues of ,4 and in
tlrt' seconcl s('ttiorl \\'c :r ,,\r - ' ..' r-,riiial ecllt.Itiotls h ,Mlz (R.) arrd lve
( ()lt)lln[e t]Lc lt,trl logzit'it1rtr, - -: - -,. 1..r,- irliitfi('eS. Tlro riCxt thc'orelnS are

1 1rr, rrLain r('slllt,ii of tltis s('( l i( )li.

Tlreorenr 1. The efrporlerltial functicttt, ,'\.

1)Departrrrcrrt of NlatherrLatics. -ftr'lurical l-rrirersit.v of Clrr,i-Napoca. Ronttrttia.
Ovj.diu. Furdui@math. utcluj . ro , ofurdui@yahoo . com

Let A e Mz (C.) and let a. B e C be the e-igenL'alues ctf -\. The followi,ng
forrnula holds

(eo_ eB oeJ-Jc'
I:' jf a+l'
"o:l 11-.1A-o- J
[e^A * (e" - oeo)12. i.f a.: 0.
Theorem 2. The trigonometric function sin-{.
Let Ae Mz@) andleta,li e C be the tigtnialues of A. Thefollou'ir,-c
formula holds

sino- sin 6.0 o sin J ..irr o ,^. il a t' i.

sinA:1I I
cr-] a- )
[(coso)A+(sin0-ocoso,ll. il o:J.
Theorem 3. The trigonornetric function cos--1.
Let A e Mz (C) and let a, P e C be the tigent'alues of A. The followi'ng
formula holds

COS O- COS p--A +f- clc .r- -cos CT

12. ,f a+ P,
cosA: j
u- l) a- 0
(sina)A * (cCOS (a-v*o
a sino)12, ,'f a: 0.

Before p rove these tI

bheo rems we collect a ierlnra ure need in our'

Lemma 4, Let a e C. and let A e Mz (C). The followi,ng statements hold:

(o) eolz : eol2;

(b) :
sin(a12) (sina)12;

(r) cos(a12) : (coso)12;

(d) if A,B e Mz(C) commute, then eA+B - e"{eB;

(") if A,B e Mz(C) commute, then sin(A*B) : sinAcosB*cosAsinB;

(f ) i.f A,B e Mz(C) commute, then cos(A+B): cos,4cosB - sinAsinB.
Proof. The proof of this lemma can be found in [4. pp. 189 192]. tr
Now we are readl-to prove these theorems.
Proof. First we consider the case e + P. \A/e have. based on the Cayley
Hamilton Theorem, that (A - aIz)(A - 0Iz) : Oz. Let X - A - aI2. It
follows that X(X + (a 8)12) : Oz ivhich implies that X2 : (0 - a)X. This
O. FuRour, Colr.lpurrxc FUNCTIoNS oF N'IATRICES tN Mz (C)

in turn implies that X" : (P - o)"-r X, for all n ) 7. We have,

eA :e aIc*X
: eol2ex
Do vn
" /-/
-oo\-^ n).
: iL (6 - n)"-'x \
't-- r1" - nt I
/ eB-o-1 \
:e" (12+#X)
J-a\- /
/ ea-o-t \
: e' I 1., + - olr)
\' o-a
: -r
-i-4 t2'
o-0 a-3
\orv we consider the case a : P. The Cayley-Harnilton Theorem irn-
:--:'-. tliat (.A- aI)2:Oz.Let X - A-a12. It follows that,4: X +alz
'.:, - -Il : Oz.\Ve have,

atldTiir,-,:::-,-..-I-''- tr
\ot- r','t :-'.': --: - - . l---- t:It- L
Proof. Onenreti--r ,::'.:--,-.-tl--:-:.-:t-tsba-.edolttheuseof Eulerrnatri,r
t, 1 _ .
cornbined to Theorem l. \\-e ieave these details to the interested reader.
Irrstead we prove the theorenr b]-using a power series technique. First we
consider tirc case a + .3.-{s in the proof of Theorem l we have, basecl on the
Cayley Hamilton Theoreur. that (-'l -aIz)(A- 0Iz) - 02. Let X : A-alz.
It follows that X(X + (a - ,3)12): Oz which implies t'hat X2: (1j - a)X.

This in turn implies that Xn : (0 - a\n-lx, for all n ) l. We have,

sin,4 : sin(X -f aI2) : sin X cos(al2) * cos X sin(al2)
: sinX(cos a) + cos X(sin a).
On the other hand,

sinx: i,-,
L^. YJ":-
(2n + t)l

: i' -1)'(9-:t:x
(2n + 1)! '.
1 ! t ) ^r2rr-l I
\-r-1,n,t \J - u/ o
' n,:o
sin(6- o)
)o tz2rt
cosX : \-r-1)'
.L^, ^

: rz *i,- ,," * t ' ,

,-l .),'r'
l2n )l
- t2t^ -r-L-1 i,- - a)2'
' ' (zn)l. n*
' cos(P
: r.2 , ___;__-\.
ll' - "
J-a '
Putti,g ali these together we have, after simple calculations, that
sin-{: sin(J - o)'(.coso)(A l_ cos(J
\vU\u-- q/
a)- 1(A-
1 I
.l_n \ /\ - alz)-/,+ lb + L-"
aI2)l sinn
0_a ' l
_ - sin0, *
sina a sin0
- 0,srnsfr.
a-A o-b
Norv n.e consider the case a : 13.The Cayley Hamilton Theorem im-
plies that lA-aI2)2 -Oz. Let X - A-a12. lt follows that,4 : X+alz
and X2 - Oz. \Ve have,
: sin X cos(o12) * cos X sin(a/2)
: (cos a) A 1- (sin a - a cos a) 12.
O. Funorr. Cortptrt\G FL'\crIo\s oF \I.{TRICES IN M2 (C) 5

\Ve nsed that sin X : X and cos X : Iz. n

The proof of Theorem 3 'ivhich is similar to the proof of Theorem 2 is
Ieft as an exercise to the interested reader.
Corollary 5. Functions of refl,ecti,on rnatrices.
Let a,b e R.. The followinq erlualzti,es hold:

(1) :") : (,osn,/pr.) ,, *

(2) .i, (1 b\ -'i"gly (' b\
\u -u) lo2 y fiz 16 -o )'
(3) ."= (l l,): (,,"JF +E) u
Rernark 6. Obserr-e that
la t \ r- ^ lcos? sin0 \
(; -") \/ o' + b' (.in o - '
r'-r,ete cos H : :
;4,zL and sin 0 i#. Thus- any matrix of the form
(': ' ) t. a scalar multiple of a reflection matrix. The reason why the
\b a)
rnarrix ('?"1,
tll^) is called a reflection matrix is given in p. 1a].
\srnU -cosU)
Corollary 7. Let a,b,c e iR. The followi,ng statements ltold

( cosh v& { "i,h /ta \

",f ? sinh v6r _ t/bc I ,t bc>0,
::):l /
\ m
/ "ot J --il' -+=
cosh v47

l, .-\
t1n u--6" \
t/-bc l rt bc<0.
", 1
c sin t/-bc
_ t/-b,
t \ -+
/_br /I
Corollar5r E. Iit ,:.:,,.c 1?. The fol,louti,ng statements hold

/ .i,, n lbc a"o.ori,

Jbr""-- v47\
I, "os
vG I ifbc>o.
.*(::):l // \;.o.
sirr a cosir

r - 6r'
cos a sinh ,/
-bc \
- - -
r -il - sinacosh l-fr
t \*,=cosa:inh )

Corollary 9. Let a, b, c € IR.. The followi,ng statements hold

I ( ;H.::; f"-:'::7,*) i,f bc > 0,

.".(: 2):

lH;:t,:-J; r;.:;#=)

Corollary Functi,ons of rotation rnatrices.

Let a,b e IR. The following erlualities hold:
la -b\
(1) o):""("?"?
"(, \sin b cosb )'
/o, -f\ _
(2) t'' (o (sinacoshb -cososinhb\.
o ): \"r.osinhb .;;;;.h; /'
/ o -a\ / cos acosh b sin a sinh b\
(3) tot(, o): (-.ino.inna ."."."rr.a/
Remark 11. Observe that
lo -b) _ (coso -sina)
(a u/ G +b, \srno coso/ .

where cos o : and sin * : Thus. any matrix of the form

ffi ffi.
la "
I ; -a\I is a scalar multiple of a rotation matrix. The reason why the

\D a)
matrix (:::: -^::":) is catlecl a roration matrix is given in [4. probtem
\slno cosa )
1.61, p. 31].

Remark 12. We mention that if r € IR and A e Mz (R) the calculation of

the matrix functions eA", sinlAr) and cos(Ar) has been done, by a completely
different method, in [4, Appendix A, pp. 354-358].
If / is a function which has the N{aclaurin series /(z) : i frPr",
lrl < R, where -R e A e Mz (C) is such that its
(0, oo] and eigenvalues a and
B verify the conditions lal < R and l0l < ,8, then
( f(a)-f_@).0*"f(il-Qf@)r,. ir 1+o'
f@):1'a-a a_ tl -
t /(")a + (f (a) - of'(a))12, if a: 0.
For a proof of this remarkable formula, which is based on the calculation of
the nth power of a square matrix of order two the reader is referred to [4,
Theorem 4.7, p. 1941.
O. Funour. Cor,lpurrNc FUNCTToNS oF MATRICES IN -Mz (C) 7

We mention that formula (1) holds in a more general case: when (t + 0

the formula holds for complex value functions / defined on spec(,A) : {a.3}.
while if n : .J the formula is valid if / is differentiabie on spec(A) : {a} (see
[2. Notes p. 221] and [3, P.11, Sect. 6.1]).


Let A e ,A4z (R). We say that B e Mz (R) is a real logarithm of A

if eB : A (see [1, p. 718]). In this section we solve h Mz (R.) various
exponential matrix equations and hence we determine the real logarithm of
special real matrices.

Theorem 13. Tuo etponential equat'ions.

(a) Let I e Mz (lR). The soluti.on of the equat'ion eA : tr12, where
rt lK 1LSgtllen0a

-4:p(tn1r1-t)fr1ln(r(-kr,t)k) \p-,
\ -r" )'
where P e GLz (R) and k is an euen'integer i,f r > 0 and an odd i.nteger if
r<o' / \
(I) Tlreequat'ion"': (1 X)u/ ,wherer,aelR, does nothauesolutions
i,n M2 (R).
Proof . (a) If a,6 are the real eigenvalues of A, then there exists P e GL2 (R)
such that

-\:P(7 9)p-' o, A-P(1 1)p-'

B)- '\0 o)-
..-, : --:-. : -, r.'hether the eigenvalues of A are distinct or not (see [4. Theorem
l- : I:-.-,:.'>:l-at
-. - -- |

and this implies that

lg 0 \
(.; .o):'r''
Thus, eo:&: o which impliesthat z ) 0 and a: g: lnr. This shows
tb.at A: (lnr)12.
If the eigenvalues of A are a*i,fl and a -i0, a e IR. and B € lR*, then
there exists P e GLz (lR) such that (see [4, Theorem 2.10, p. 79])

t: p( o^ o) p-t.
\-p a)

This implies that

/a ll\
: *b
and it follows. based on part (1) of Corollary 10, that
sinJ) :.rtc.
\- srn l, cos ) U

Thus,eocos0:r ande*sinp:0. It followsthat sinB:0+ 0- kn,

k e Z. The first equation implies that ea : r(-1)k + a - in(r(-1)k), where
k is an even integer when r > 0 and k is an odd integer when r < 0.
(b) Passing to determinants we get that
(' 1\
" :- a"t
det eA ""'\u o)
+ er'(A) : -y2,
which is impossible over IR.. D

Remark 14. Part (a) of Theorem 13 states that the real logarithur of the
nratrix rI2 is given by

: P(t"('(-,-])*' kr \ --
ln(r(-l\k))' '
where P e GL2 (lR) and k is an even integer if r ) 0 and an odd integer if
r < 0. Part (b) of the theorem shows that the -at.i* (l € R'
\U 'o)' ''9
does not have a real logarithm.
We mention that if A € Mz@) the equation eA : zI2, where
z € C*, has been studied in [4, Problem 4.36, p. 219]. Also, the trigonometric
equations sin A : 12 and cos A : 12 over the real square matrices have been
studied in Appendix B of [4]. 'Ihe equations sinA : rI2 and cosA: rI2,
with r e IR, can be solved similarly to the technique given in Appendix B of
Theorem 15. The real logari.thn'r of scalar rnultiple of rotation rna-
(a) Let a e Mz (R). The soluti,on of the equat'ion

:(; #) , g€R*,
i,s gi,uen by

f q1-r1ky.. - (2k + t);

A: | " 7r
+ t;
\(zr ; In((-1)fre)
where k e Z 'is an euen i,nteger when'y>0 and an odd i,nteger when y < 0.

O, FuRnuI, Cotlpuumc FUNCTIoNS oF i\.{ATRICES rN ,AZz (C)

(b) Let A e Mz (R) and let r,y e R*. The soluti'on of the equati'on
"o:('\v x)
i,s gr,uen by
tnzT? -lo + 2kr)\
^ (rn o+2ktr
o:\ 11r/Pafr)'
where k e Z andO e (0, 2r) i,s such that
ct.,sd: -L nnfl -sinl: -L.
t/", + y' t/*' + a'
- "\
A: (: d)
Proof. (a) Let f). s,,r." A commutes witn (f get that
\r \e i )*"
A: (? -b)
a/ rv" have, based on
part (1) of Corollary 10, that

", ;:,:') (:nl ;,) : (l

and it follows that eocosb:0 and eosinb: g. The first equation implies
that b : (2k+l)5, k e Z, and the second equation shows that e" : (-7)kA.
Thus, a : ln((-1)kg), where k is an even integer when y > 0 and an odd
(b) Let A : (: f). s,r.. A commutes *itr, (j -') *"
x/ get that
\c d/ \v
/^ r.\
: : \\'e have, based on part (1) of Corollary 10. that
Q )
-+ ( _"
\0 /

.. (.?= |
'\-inb -'ilb\ _ f, -v)
6r-'i -- -,:----.-. -: a: at --=: : : ai,i e'srni : ;,. This impLie: that e2o : r2*A2
tr: -#.
- c: -,- 1 .- - - ,--,i . .:'-,- ---=+- al,l si: Let d e (OrZr1
be such thar cos , :------- u-,-'...1: :-. r, rlil'# ,nr,
\^ - \- -'

fsinb : sind
/.=1 l', :, - ]i:: or b: r - 0 +2kr,
[cosb:coS0 ltr-n-2n- or b:-012rur,
where k,ne Z. Stn.cer,A T 0 these ca-=es imply that b:0*2kr,ke Z. a

Rernark 16. Part (a) of Theorem 15 shows that the real logarithm of the
real mat,rix [ " -,,\] i. gir-en b1-
\Y t)/
r,. (o -y\ :- (tn-1.t-tf y) - (2k * l) ro)
"' \s o ) \rz,t + t1; in((-1)kil )'
10 ARrrcoln

where k e Z is an even integer when y > 0 and an odd integer u''hen g ( 0.

Part (b) of the theorem shows that the real logarithm cif the rnatrix
(x -v) r
\v r/
n ( * -y\ : (tt, 1/r2 + s2 -\0 -r 2A'n) \
"'\y ,/:\ o+2kr tnF+i)'
where k e Z and d e (0,2tr) issuchthat cos0: -"{rr*0, ancl sin0 - ----L6+rr.

Corollary L7. The real logarithrn of a rotation rnatrir.

Let a € (0.2n) \ {i,", f}. fne solut'ion,, i,n Mz(R), ,/ th,e r:quation

("g..i -sin,r\
"r: \srn 0 "or., /
ts g'iuen by
A: (a + 2k7Tt (0. -^l) . k e z.
\r u/
The previous corollary shows that the real logarithm of the rotation
matrix (cosn - sino) '.
\srno cosG /

(:il; ;l:') - 'r"^)
(o * (? l,')
t' e z
Theorern 18. The real logarithm. of a circulant rnatrir.
Let A e Mz (R) ond let r,y € R, U 10. The equat'ion

: (' Y\
has solut'ions'in Mz(R) i/ and only if r > 0 and -r < y < r and in this
case the soluti,on ts gi,uen by

(t,,r/ir] t,,\P, \
'- h\F, ntP:u,)
Proof. Sin.e A commutes,irt, (' 1) *" set rhat .4: (i '). ,,0 nur".
\v r) " \b ,)
based on corollarY 7' that^
e' : e'^ (slnh sinhb\
b coshb) '

It follows that
, :r)o
"'' + 1""*o:r-l-y>0.
lpa "b --"-' :Y \e'[z-a:r-y>0.
("' 2
O. Funour, CoMpurtt{c FUNCTIoNS oF MATRICES IN ,AZz (C)

Thus, -r < A ( r and a calculation shows that o :Int/*2 -V und

b:rn,H tr

Corollary L9. Let A e Mz(lR) and let g e IR, y ) -1. fne solution of the
: (';'
'o ,t. r)
l.* q1r.r rt bLl

A:tnJfr-(+ i)
Proof. In Theorem 18 we let r: y -F 1 > 0. D

Theorem 20. Letr,lJ e R bc such thatry <0. The solut'ion of th,e equati,on

"': (: 6)
r,s gluen by
/ "- Tx.-. -\
hJ-rY *;l-:(zk*')l
^-l[ aL,f
'^- l,J-'a l'
\ 2u-_.12a,-1)
'r' )
where k > 0 i,s an integer.

Proof. Let,4: (: 3) . rr,u,. Sirrce -{ col,mutes *r,r, (! 6) *"

have that l: (o O). Onr.r.-e that Corollalr-7 irnplies that br < 0 and
\r' a/
we get that

//-h-\cos J-bc -+ sin y'-bc'l

"'[#sinr/-bc ';';: J-* _: f,(; ;)

which implies that

f.,F-0 r,/-bc : r. e)
\" @sin
sin ,/ - bc : s.
1"" i-tu
The first equation inrplies that J-6r': tQt * 1), where k > 0 is an integer.
It foliows tnat e"fr(-1)k : r and eal+fre\o : a.These two equations
imply that e2" - -:xA =) a : ln 1/-ry. Dividing the last two equations in (2)
t2 ARrroor-p

weget that | : t +b:1". Ontheother hand,bc: -(tQn 1r))'and it

follows that c2 : -v (;Qn 1 r))'. This implies that c : t|1[-yen * t1

and b: - Gtr)' G;r[+ek + 1)) : +; yffoer *

i,: ! rt
Theorem 2L, The real logarithm of tri,angular matrices.
Let r,g e R*. The equat'ion
^o- (* v\
":\o *)
has solut'ions i,n Mz (R) i/ and only if r > 0 and i.n thi's ca,se th,e solut'ion 'is
gi,uen by
o-lnro lnr)'
: (' q\ since A commures *itn (l
-Proof. Let A-
\c d) \0 r/ ') *" eet that
A: (2 a)
1) ," this case both eigenvalues of A are equal to o,. It follows,
based on Theorem 1, that

- ("o
,o be'\ 1r u\
0 e'/-\0 r)'
Thus, ea :t) 0 and beo : A. This impiies that o, : lnr and U : *. tr

Remark 22. Thepreceding theorem shows that the -utri* (f f) no. u

real logarithm if and only if r > 0.

\u r)
Theorem 23. The real logarithm of syrnrnetri,c rnatrices.
Let X € Mz(R) be a symmetric matrir such that X f aI2, a e IR.*.
The equati,on eA : X has solut'ions i,n Mz (R) i/ and only i,f Tt(A) )> 0 and
det A > 0 and i,n thi,s case the solution i,s gi,uen by

.4: l,]r)r-,\z
- l,)2, , )'1'|z - ]zl')rrr.

where )r, )z are the e'igenualues of X.

Proof. Since X * alz and ,4 commutes witli X we get, based on Theorem
1.1 in [4, p. 15], that A: aX * 0Iz, for some a,p e 1R.. The equation
eA : X implies that eox+glz - X + etX : e-1X. Since X is symmetric
and X * alz, a € IR*, we know that X has real distinct eigenvalues (see
[4, Theorem 2.5, p. 73]). Let Ar,,\z be the eigenvalues of X. Observe that,
since A commutes with X, A is also symmetric and it has real eigenvalues.
If p1 and p2 are the eigenvalues of A, then {e+l,euz} : {,\1,)z}, Thus,
O. FuRour, CorrlpuuNc FUNCTIoNS oF MATRICES rN Mz (C) 13

both eigenvalues of X should be positive real numbers, i.e., TY(A) ) 0 and

det,4 > 0. The equation eoX : e-A X combined to Theorem 1 show that
a)1ed^z - a,\2eolr
12 : s-0 Y.
a),1- a),2
This implies tliat
e^lr - eo), _- );g^^z - )2g^)r -
)r-), x-r lz:'- _R., '\
A1 -4
and, since X f aI2, a + 0, we have that
- 1 )1eo)z - )2eotrr
)r-)z "a.\z
)r-)z -0.
- \
), -ln )c
- l2eolt :0 +
Itfollows that )1ea)z )r -)z
the other hand,
eaAt _ eo Z

:ln )r-lz
.\, .\"

(+)^'-^' - (f )^' "

: .)r
/ )r \ ^r-lz
_ )rln)z-.\zln)r
A: ax + sI2-
)r - ln )z x + I!]?-*Elrr,
"rd the theorem is proved.
Remark 24. Theorem 23 shows that a symmetric matrix which is not a
sr:a--:-: :r,,rltiple of the identity matrix has a real logarithm if and only if its
eisr:-'..r--.re= are both positive real numbers and in this case the real logarithm
iS g-',-t:- :,-.' -,-,:nlffla (3).

f1] D \. Berr..:=::-. ,l-lrrirt.z' Jlathernatics. Theory, Facts and Forrnulas, Princeton Uni-
versitl Pr, -:. I ,

f2] S. R. Garcia anrr R. .\ Horn, .4 Second Cou,rse i.n Linear Algebra, Cambridge Uni-
versit;'Press. li'11.
[3] R. A. Horn and C', H. J,:hnson, Topics in Matri,u Analgsi,s, Cambridge University
Press, 1994.
[ ] V. Pop and O, Furdui. -i4uare -\Ialrtces of Order Two. The.ory, Appli,catzons, and,
Problents, Springer. 2017,
t4 ARrrcole

Stirling type formulas

IoN-DpNys CroRocaRu')

Abstract. In this paper we prove two Stirling type formulas:

9k 'nz z
r,2 n s
il u+il- J2"
i+j<n;i,j€N "
' ^ nn'-i.4"2'"
f[ ti r jt - aH
i..i=1 e 2

*.here gr is the Glaisher-Kinkelin constant.

I(eywords: Stirling formula, Glaisher-Kinkelin constant.
MSC: Primary 40A25; Secondary 41A60.

1. INrRooucrroN
In this paper the notation and notion used are standard. in particular
N : {1,2, ...} is the set of all natural numbers.
Definition l. Let (b")rerv be a sequence of real numbers wi,th the property
that lrls € N szch that b, I O,Yn ) no. We ui,ll say that the sequence of
real numbers (a,,)rex is equivalent wi,th (b,),,e x and we write an - bn i,f and
onlu if lim 9o
" " on
: .l-
From the well-known result that if a sequence of real numbers is con-
vergent then every subsequence is convergent and has the same limit, we

Proposition 2. If an - bn then a2na1 - b2n+r.

We need the following two results.

Proposition 3. Let "f , (0, oo) -+ IR. be a functi,on. The following formula

f (i.+ il:i(k - 1) f (k)

itj <nri.j €N &:1
Proof. \\'e have:
n-1 n-L n-2
f (i,+i) : I (-r(c+i)) :I,ro+i) +I re+i)
2fj<n;j,j€N i:1. j= t 1- |

1) N{ihai Viteazul Secondary Nr. 29 Constanla. Str. Cigmelei. Nr. 13. Constan-
ta and Department of Nlathematics, Ovidius University of Constanla. Bd. \,Iamaia I24.
900527 Constanla. Romania. cidenys@gnail . com

+ + tj:r f(("- 1) +:) : f (2) +2f (3) +...+ (n-t) f (n)
:I(,k-1) f (k).

Proposition 4. Let "f , (0, co) -+ R be a functi,on. The following formula

n n 2nl-L

Irtn+7) :Itr
iJ:t k-7
1)/(k) + t
k:n] l
(2n-t7 k)f (k).

Proof. We have

ro+r) :I (t
rL r, /, \ ,

f (i+r) ) :Irrr+i)

j-t j:l
+i) +
i, ,:1 \J:1 /
-1 ln+ j): (2) +2f (3) +...+ (,
+>_f f - 1) f (n) +nf (n+t)

+(n-1)/(n+ 2)+...+f(n+n): t,u-1)/(k)*Ir, D f @+k+ 1).

A-1 &:0
Changing n + k + 1 : i, we get
n 2ntl
It"-k)f (n+k+1) : D Q"+1-i,)f (i,)
,k:0 i:n11.
and so
n rL 2n*7

L t tn+ i) :Ir* A'- 1

- 1)/(k) + t
k:nI l
(2n+r - k) f (k).

We recall two well-known results. Their proofs can be found, for exam-
ple, in [7].
The Glaisher-I(inkelin theorern. The sequence (r,,,)r.er,l wltere

11 .22...Tln
n2,n,7 n2
rt-Tt2-r rz . g- a
is conuergent and i,ts limit lim r, : gr is called the Glai,sher-Ki.nkeli,n
16 ARrrcor-B

The Stirling formula. The follouing formula holds

;1, - Jr;; X
\['e will use these results to obtain the tu,o Stirling type formu]as stated
in the Abstract.


Proposition 5. The followi,ng formula holds

n2-n 2-
n2 5

fI t,*r)- +
'Iifu' '/zi "*-'
Proof. Using Proposition 3, we get

rn II t, + i) : I rn(i + il :ifn- l)rn k

i*j<n: i+j<n; k:l
i,j€N t,i€N

: i ktnk -!r,4, : 1n
;1 o- - h. II k : r,H
,b:1 k:1 ft:1 k:7

fl r*

t,i€N l[r
From Glaist theorem and Stirling's fbrmula we obtain

il (i+i)-'ffi:hffi ,r2-,r-l ,r2 n2-n-5

I f.i <n:i..le N v att ?n


Proposition 6. The followi,ng formula holds

1'o -2'/, !#:
I.-D. CIoRoGARU, STIRLINc typE FoRI\4Lrt,AS 77

Proof. Using Proposition 2 in Stirling's formula, we will have

2nll .,2n+l
f[ r- l-'2" Y!++ An+t
2n*1 ll k
From tlre equality II k: +- and Stirling's formula we obtain
k:n*r ll k
(2n * 112n+1 (2n+ 1,2n'1.,12+-
2n11. J-2- T.\2rJ
,k:n* 1
nn . en+L

,n .
Since (1 +*)'"11 -eurd f + - r'f. \i'eger


Proposition 7. The following formula holds (l - +)----- - {3 .e".
Proof. Indeed, we have
(2n + 1)2
J*\-##- _ i_ [(,.;r)"1

l, -l'.. i)^,) :,'B(('*#)"-r )@#

L 1, ))

(O+*)'" _r\ (2"+r)2:

Hm n((r+*)'" _,\
;5'J\ , -') * , -')--W-
18 ARrrcor,p

: lim
\ )
Flom the L'Hospital rule we have
,i* t (l+*li _r\:1u,,,
e i

i'If2.\ " -L):ril,i --1 r


Inrl+/J_l lnfl .J)

rr ,. t'/ f, - 1 , - |' I- ,. ln(1 *r)-r
:-llm rn(r+r)_1 :-[lm
2 i1O I 2 r_+O

:1rr,,, l+r-1:-1.
2 r-+0 21 4
From the characterization of the limit of a function at a point with
r,* , tGSd

sequences, we deduce that e-4. so that

\2n-l t2
/ 7 \ -- i2n+r)2 1
4:e,, -_E_
It+- | rue 1n 4 4n.
\ 2n)
Since ea"
- 1. we obtail (t + #)
2 - i/e3.e'. !
Proposition 8. The following formula holds
2T!t e2n2-3n ,**'+*,
II kk _ 2e,:/i .___________
k:n*7 e 4

Proof. Using Proposition 2 in Glaisher-Kinkelin formula, we have

"n' ^ _,
rr - 9t, . (2n + 1)
u#f *'**i - r-e'+t .


2n*t 'fi'no
Then from the equality ll lxk : J=J-- we deduce that
x:*t ,!,

2n*L Qn+Lt2 2n { r+ I r2nt t\2 l2n*l t2 , 2n+1 , r

gx. (2n-1)------- 2 t2 . e- T (2n + l)----a-t--T--r n
L n- n2 , n
n 2T2-t2.e-
I (n-t)(3n*t]

en-rf 2n+r r / 1\--r-
-n (, *
(2n)----z-* 2
n2 ,,
, I (n*1)(3n-1)
(,* *)*r
n,-TazTu.c 4

I.-D. Ctonoca.Ru, SIIRI-IItG TypE FoRMULAS

22ri2-3n. ]3 ., *-'; -, (, * 1+
1 \'++
,l 2")

From Proposition 7 and (r +#)"+* - ,/i *" obtain

r - 22n2+s,*# . n**!+r . {@ .en . rt
II " e'-t-..l-llr-]L
k:n*l 7

22n: - 3n . ,''r' , "; +l

- ,-

Proposition 9. The fotlowi,ng formula hold- (l * *)n"'-Snr'3 - e2.e2n.

Proof. We have

t1+-lI 1 an2+sn+t --)'(,-i)'.'
\ 2")
Iim --en;+- : Iim
t, ) ''G

By Proposition 7,
\Zn+tl2 12

('- *)....-
---or-) |

Proposition 10. The followi,ng fonnula holds

Proof. From Stirling's Theorem. see for example [7, Theorem 5], we have

f{ r : J2r; .'j .",-- . "*".

20 Anrrcor,r,;

where lErl < u{ub,Vn € N. Changing nto2nf 1, we get

(2n -t t12"+t
ff t' : \/,;e;TT . . sR2n+7,
k:t "rr*E
wlrere lRz,,-tl S
*#;,f.Vn € N. Using these rerlations) we u,ill obtain

2rtt7 f{ t'
j#-: 2r (2n *
(2n + 1)2"+1
. e 12\2n+1) ett2n+t

ilk: ":D
l2n ' gttn

- C",r!
(2n t- t)2n-s . n-l'-i)
Qn)'"-i (' . *)'"' '
'-@+i) .an
en+| . eTr;- t2t2bt' ari-1 . e12n- t2t2*1 '

22'-i./r*a\2"/ '.rntl
::'on, \
where {Yn, : sRzn+r -8 . Then we deduce that
/ 2n*r 1 2n*r 2lz,t'))tzn vrl ( , *: )
"' . ,,(n- t)(2rt*t)
where gn: - r(2n-t1)(Rzn+t-R"). Since 16r] < #,
a2*+r we will have
lR^1.(2n+ 1) < W, and when n -) oo, we have that R, .(2n + 1) -+ 0.
Irr the salrle way, Rzn+r ' (2n -t1) -+ 0, so 13, - 1. Also, e# - gl'
using all that and Proposition 9, we get "#.
/ 2:!t 1zn+t 2Qr. i)rzn ri1 . t2. n(n I l)(2rr -1)
n- l tl2n
' ' "S
\o -r*, ,/ "r
zer/2 24'n2
+5rr . rr2n2 +:ln+l

w "2n,2

Proposition 11. The folloui,ng formu,la holds

;1 ,n * i) - *F"*o
' ""'-# 'n'*n

i'j:l 2



Proof. Using Proposition 4 rvc obtain l" fl (i+i):


ln(z + 7) :
(n kk) (-+i'n)'"*'
: i(k - 1)lnk r T rr*+ 1 - k)lnk : ln \ A-:1

k:1 k--n*l
-!,-) (-Tj,--)
and hence II (i +j) :
From Glaisher-Kinkelin, Stirling, Propositions 8 and 10, we obtain
I t ,,, \ / zrr/2 21n2+5n . n'2n2 -3rtt1
\0r.,22 |,-w ----er;rt, -
22n2+3n .n**+*,
(oa T) (*'* e

;1 t ,o * j)
- ,!=i. n"'- l'#"
vLttt e 2

Acknowledgments. The author tharrks Prof. Univ. Dr. Dumitru
Popa for his helpful suggestions that make possible to obtain these formulas.

t1l Gh. Gussi, O. Stdnaqild, T. Stoica, Matematicii, Iv[anual pentru clasa a XI-a, Editura
Didactici qi Pedagogicd, Bucureqti, 1996.
121 D. Popa, Erercilii de analizd matematicd". Biblioteca Societalii de $tiinle l\'Iatematice
din Romdnia, Editura 1\4ira, Bucureqti 2007.
t3l D. Popa, The Euler-Maclaurin summation forrnula for functions of class C3, Gazeta
Matemati,cd. Seria A, Nr. 3 4, 2016. 7-12.
22 Asrrcor,o

Some generalizations and refinements of the RHS of

Gerretsen inequality
N{lnrus DRXcaNt), NBcuLl.r Sramcru2)
Abstract. This paper presents some generalizations of the RHS of the
Gerretsen inequaiity.
Keywords: Blundon inequality, Gerretsen inequality, geometric inequali-
ties, the best constant.
MSC: Primary 51M16; Secondary 26D05.

In [5] appears the inequality a2 + b2 * c2 < 8R2 + 4r2 dte to J.C.

In [6] L. Panaitopoi proves that the inequality of Gerretsen is the best
if we suppose that
a2 + b2 * c2 < aR2 + pRr. + 1r2,
where a,0,1€ lR. and 0 :0.
In [7] the same result as in [6] is shov,.n for a, 13,1 e IR and p > 0.
In [8] R.A. Satnoianu gave the following gerrera,lization of Gerretsen
un +bn *c' < 2n-tpn +z'(!+; _ 2n- l).". for anv n ) 0.
The purpose of this article is to give a proof of Satnoianu's inequality in the
case ?z : 6, i.e.,
o6 + b6 * c6 < 128 . ,R6 - 3oo8 .16,
then we prove that ttris inequality is the best if we suppose that
a6+ b6 * c6 < a1R6 + a2Bbr+ ... + a6Rr5 * a7r6,
where (tttc\2,...,(17 € ]R. and (\2,a3,...,oZ ) 0.
Also we give some refinements for this inequality and we shall prove
these refinements are the best of their type.
Theorem 1 (fundamental inequality of Blundon's inequality). For any tri-
angle ABC the i,nerlualitzes s1 ( s ( s2 ltold,, where s7)s2 represent the
semi,peri,meter of two ,isosceles tri,angles ArBtCt a,nd A2B2C2, wh,ich haue
the same c'ircumrad,,iu,s R, and ,inrad,ius r as the triangle ABC and the std,es
at : 2l(Rt r - d,)(R - r * d), bt : ct : + r j,
a2:2J(R* r * d)(R - r - d), b2: c2: t/zA1A +, + a7,
where d : V@-Rr.
A proof of this theorem is given in [3].
1) Mircea celB5trin Highschool, Bucuregti, Romania,
2) G.p. Palade Secondary School, Buzdu,
Romania, marius.dragar2OO5@yahoo.con

N,{. DnXceN, N. SraNcru. THs RHS or GeRRorsEN INEeUALITY

Theorem 2 (some minimal and maximal bounds for the sum a6 + b6 + c6).
In any tri,angle ABC is true the double tnequali,ty
24(R+r-d)3 [+(n- r+d)3 +E']
<16+b6+c6 <211R-r-Fd)3 [4(E -r-a)3+n3] . (1)

Proof. If we replace r: a2. lJ:b2. :: c2 in the identity

fc]'c r'' : jJr,,. - I,.i : (f r'- !rr) ,

\ cvc cyc /
then r.",e obtain

\-,': .,).-,.---\- -itn,-f a,",)


- l' . c]'c

: (t,)'- +\ab(X,) '*u,o,Lo- (I,o)'

\ .r'. / ,y, \ft ) '.r. \ ... /
: s4 - 2r(4R+Tr)s2 + (4R+r)2r2,
F(s) :/(")+e(s)h(s).
Since h'(s) :4s(s2-4Rr-7r2) and s2 > > l6Rr -5r2 > 4Rr *7r2.
itresults that h is increasing on [s1, s2].
Also / and g are increasing on lsr, sz]. Hence, F is increasing on ["r,
and then F(rr) < f.(") S F(s2) or
a!+u6r+"f < o6+bG *c6< (2)
If ,*'e replace a1,b1,ct,a2,b2,c2 from Theorem 1 in (2) we obtain (1). !
Theorem 3 (some maximal bound for the sum a6 + b6 + c6). In any triangle
ABC holds the i.nequali,ty
a6 +bG *c6 < t28R6 -300816. (3)
Proof. From (2) it follows that to prove (3) it will be sufficient to prove that
16(.R+ r+d,)s lt@-, - d)3 +E'] < 128-R6 - 300816. (4)

If we put tr : i,R then the inequality ( ) can be written as

to (r+ s 1/az - n)' ("- ; 1/*'-2,)' * "'] < L2Br6 -2008, vr ) 2,
24 AR'rrcor,p

which is successively equivalent to

(6415 + 64ra + 4Br3 - 204812 -t - JBq\[r@ -
2b60r 2)
{ 64rG * 4Br5 - 206413 -t 406g12 - 1920r - 2944,

16(r - 2)@ra *
* 2T* - T4r + t2)\/r@ - 2)
< 16(r - 2)(415 I Bra * r9r3 - g7r2 f 106r + g2),

(4ra + t2r3 + 2Tr2 - T4r-t t2)\/,4r - 2)

I 4r5 * Bra I tgr3 - gtr2 * 1062 * 92,

4818 +24017 +8163.6 +TB0r5 +24tra -t|T2rs -920412 +lgT2r*8464 > 0.

Y r )> 2, which is true since

81616 _ 92o4rz:81612 (. _ #) > 81612(."4 _ 16) > 0

\ 816 ) -

78025 - r712r3:780.rr (r, _ g) > 780r3(r2 4) > o, vt.z2.

\ 780)-
Theorem 4 (the best maximal bound for thc sum a6+b6+co is 12g,R6-300g16). I/
(1r,a2,...,a7 € R. ond a2,a3,... ,(t7 > 0 with the property that the i,nequali,ty
(5) is true i,n euery tri,angle ABC
a6 +bo +c6 I atB6 +a2R5rla3R4r'+onr3r3 +o5R2ra +a6Rr5 +a7r6, (b)
then we haue the i,nequali,ty
a1 R6 * a2 R5 r+ a3 Ra 12 + a a R3 r3 + a5 R2 ra + a6 Rrs * az rG > 128R6 - 300816
i,n any tri.angle ABC.
Proof. In the case of equilateral triangle. from (5) we ha'".e that
2uol + 2'o, + ... t 2a6 + a7 > 222. (6)
If we consider the case of isosceles triangle ABC, b: c: 1, a:0, B :
r:0, then by (5) we deduce
a1 ) 128. (7
Taking into account R> 2r. (6) and (7), we successively obtain that
(a1-128)rB6+ azRsr+ asRar2+ a4R3r3+ a.sR2r4* a6Rrt'+ (crz+soog)r6
> [(o, -728)26 + o2. 25 a a3. 2a * aa. 23 + o.s. 22 + aa. 2t a7+3008]r6 > 0.
\I. Dn.c.cA,ri. N. Sr.qNcIrr. Tuo RHS or GsRRsrsEN INEQUALITY 25


a1R6l .,2Rlrl a3R4r2+ aaRsr3+ a5R2ra* a6Rr5+ a7r6 ) 728R6 -300816.


Theorem 5 (the best constant for certain Gerretsen type inequality)- The
best real constant k such that the inequali'ty
o6 + b6 * c6 < 128R6 * kRrs - (2k + 3008)16 (8)

'is trure 'in euery trtangle ABC i,s h = -5269-275'

Proof.Inequalitl- (1) yields that if (8) is true in any triangle ABC, then
16(E+r+d)3[4(R -r-d,)3 +]?'l < 128R6 +kRrt'-(2k+3008)16 (9)
is true in an1. triangle ABC. The ineclualit\' (9) is successively equivalent to
16(n+ r+ d)3[4(.R - r- d)3+ /?3]- 128R6+300816 < krs(R-2r),
tr f

-i ftu,;r- t* /*z-n1314(r-1-v,r'2, i't- r3l- 12816+30081 < k'

16(-415 - Bira - - 1913 * 9112 106r - 92)

+16"F - Z* l4ra -t l2r3 + 2712 - 71r -t 12) < k.

so the best constant is the maximunr of the function ft,1.2. x) -+ R..

:f Jr) : i6(-4r5 - Bra - 7913 f l9r2 - 106r - 92)

+rc.,81m(ua +7213 +2712 -7)t +12).

i.e.. kr : Tgf "fr(r) = -5269.275, tr

Rernark 1. The best integer constant for u,hich the inequality (8) is true in
following inequalitl'
a6+b6 *c6 <128R.6 -5269Rr5 *753015.
Theorem 6. In any triangle ABC holds th,e followi'ng i'nerluali'ty
o6 + b6 + cG l-128R6 * ARr5 - (2k + 3008)16. for eaclr,,k e [k1, oo),
where k1 represents the best constant for the i'nequali'ty (8) '
Proof. If we consider the increasing function
F:lk1,oo) +iR, F(,k) :128R6-300816 +kr5(R-2'), thenby (8) we
infer that
o6+b6*c6<r(kr) <r'(k) (10)

for any k> h. D

26 ARnlcor.p

Remark 2. If we take k : 0, then by (10) we obtain a6 + b6 * c6 < F(k),

i.e., a refinement of inequality (3).
A generalization of Theorems 5 and 6. For n e {1,2,3,4,5}, find the
best real constant k such that the inequality
a6 + bG f c6 < L28R6 1 7rpnr6-n - (2*k * 3008)16 (11)
is true 'i,n any triangle ABC.
Proof . By (t) we have that if the inequality (11) is true in any triangle ABC,
t6(R + r + d13 1+1a - r - a;3 + R3l < 128,86 a 7,
pn r6-n
- (2" k + 9008)16 ( 12)

is true in any triangle ABC.If we denote fr: {,r th"r, the inequality (12)
becomes successively
rc(r+t+t/M)eV@-t-t/7 -u)3 + r3l-para +zooa <.k,Yr)2,
rn -2n
r6(4rs -8r4-1913+ t06r-9\+(aaa a12as a27a2 -74r-ll2)l6t/@ -2"
9r:r2 -l-
rn 2 . 2+... + r. 2n-2 + 2n-7
xtn-t +
So, the best constant is the maximum of the functions fn i l2,oo) -+ R,
r, _ \ _ 16(-4;15 - 8r4 - 19u3 + gL* - to6r -92) + i6(4ra + 1273 +22.12 - ar + D) tff - 7 za

lf n:1, then we find the best constant from Theorem 5.

lf n:2. then (using Wolfram Alpha) we find k2- -1297.57.
If n : 3, then (using \\bifram Alpha) we find ks > -479.402.
If n :4, then (using Wolfram Alpha) we find k+ -96.
If n:5, then (using Wolfram Alpha)
we find ks 0. tr
1) The best integer constant for (11) in the case n:2 is kz: -7297.
so the inequality
a6 +b6 * c6 < r2BR6 - r29TR2ra * 218016 (13)
is true in any triangle ABC;
2) The best integer constant for (11) in the case n:3 is kz: -419.
therefore the inequality
+ b6 * c6 < 128.86 - 4tgRsr} * J44rG
aG (14)
is true in any triangle ABC;
3) The best integer constant for (11) in the case n : 4 is kt : -96.
thus the inequality
ao + b6 * c6 < r28R6 - 96RarB - t4z2r6 (15)
\1. Dn,lcax. N. St'.tNcn.r, Tuo R,HS op GpRRersEN INtrQUALITY

is true in any triangle ABC:

4) AIso we proved that the inequality

a6 +b6 + c6 ( 128R6 - 5269Rr5 * 753016 (16)

is true in any triangle ABC;

5) If we compare the inequalities (13), (14), (15) and (16) we observe
that there are not two of them u,ith right-hand side u(B,r) and u(R,r),
respectively, such that u(-R,r) < u(fi. r) for each !ar.
Theorem 7. If n e. {7,2,3,4,5}, then in any trtangle ABC holds the fol-
lowing i,nequaltty

o6 +bo *c6 < +kRnr6-" - (2'A- 3008)16 for anyke lkr,n),


where kn represents the best constant for tl)).

Proof. \Ve consider the function G: Lk". -) -:-.

G(k) : 128R6 - 3oo816 - l'''6 R' -'2'r'" )'

Since G is increasing function, froni (8) ri-e har-e that

a6 + b6 i co < G(k") < G(,k) for any k ) k,,.


Theorem 8. If n € {7,2,3,4.5}, at,(72,...,a7 € R,

{ir. ir. zr. l+} : {2,3,4,5,6} \ {6 - n * 1}
such, that (tit, c\it. oi:. crz+ ) 0, a6-21 | ) kn, with the property

s6 as6 a s6 ( o 1 -R6 - n z R5 r + az Ra 12 + o n R3 13 + o5 R2 ra + a6 Rr1 I a7 ru, ( r z)

then we haue that the inequahty

l,z8 R6 +k, Rn 16-

n- ( 2',k, +3008)16 < ay R6 I a2 R5r* ... * ae&rs + a7r6 1ra)
'is true r,n any trr,angleABC. where kn represents the best constant for (71).
Proof. In the case of eqr-rilaterai triangle from (17) we get

arR6+ a2?br +ci3/?lr2+alE3r3 +asR2ra+o,6Rr5 laTrG >722. (19)

In the case of isosceles triangle r.vith the sides b : c : 7, a :0, E : 'r, ,: O,

from (17) we have

cr1 ) 128. (20)
28 ARucoln

Taking into account (19) and (20), we obtain that

(a1 - aitR6-ittT ri't-r + airR6-iz*7riz-r * airftl-iz+t rnr-,
128)ft6 +
*ainR6-it+trit-r * (ao-r,+r - kn)R"r6-" * (o, + 3008 *2,'kn)r6
) [(", - 12q261rl,ir 26-i'+t * oi, ' 26-i'+t
+r,i:j 26-it+t + oi4 26-i'+t * (ao-,,+r - k,r)2" + CI7 + 3008 + 2"kn) 16
: fo,' 26 + a2 . 25 + as . 2a + at . 23 + a5 . 22 + aa . 2 * at - T2')16 ) 0,
which yields that
o1,R6+ azRt'r+... 1ai;-Rr.r'* a7r6 > 128R6 * krR,16-,, - (2,rk,-* 300g)16.

f1] W.J. Blundon. Problcm 1935, Tlte Amer'. Math,. Monthly Z8(1966). t122.
[2] w.,I. Blunrlon, Inequalities associated with triarrgle. canad,. Math. Bull. g(1g6b), 618
[3] M. Drdgan and N. starrciu, A new proof of the Blundon inequality, Rec. Mat. 19(2017).
100 104.
[4] Drdgan, I.V. Maftci, and S. Ridulescu. Ineqalitd"Ii Matematice (Ert,indrcr.t, q,i gene-
ralizdri), E.D.P., 2012.
l5l J.c. Gerretsen, ongelijkheden in rle Drichoek. Nieuu Tijdsr:hr. wisk.41(1953), 1-7.
[6] L. Panaitopol, O inegalitate geometric.i. G.Xt.-B 8Z(1982), i13 115.
t7l S. Rddulescu, N,l. DrAgan. and I.\,'. \Iaftei. Some corrsequences of W.J. Blundon,s
irrequality, G.ltf.-B 116(2011), 3 9.
l8l R..\. Satrroianu. Gettcral pos-er itrr:qualities bctri'een the sitles arrd the circunrscribefl
arrd inscribed radii rciated to the fundanrental triangle ineqrralitv, Math. Ineq. €t Appl.
5(2002). 715 751,

Olimpiada de matematicl a studentilor din sud-estul

EuroPei, SEEMOUS 2018')
G-.reRrBr N'ItNcu2), IoaNa Llica3), CoRNot-
B'[olrcaa) . TtsnRtu TnIr'5)

12tr' South Eastern European Nlathematical

Olympiad for
Abstract. 'fhe bv the Gheorghe- Asachi
Li;;;y Students. SOEITOUS 2018' *'as hosbed
27 and March 4'
Teclrnical University. taqi Ro*unia' bets'een February
probl"r-,* a,d their sol,tions as giverr by the
iv;^|i.-,;a tt " .orrip"tliio. by some of the competing stu-
authors. Solutions providecl
dents are also inclutled herc'
I{eywords: Diagorralizablc matrix' rank change of variable' integrals' se-
MSC: Primary 15A03; Secondar-v- 15A21' 26D15'

adresat5 studen{i-
versity stuclents) este o competilie anuala de matematicx.
A 12-a editie a aces-
lor din anii I gi Il ai universit[1ilor c,in sud-estul Europei'
qi 4 martie 201-8 gi a fost gSzduit5
tei competi(ii a ar-ut loc intre 27 februarie
din Iaqi' Romania' Au
de cxtre tiniversitatea Tehnic5 ..Gheorghe Asachi"
N,Iacedonia. Grecia. RornAnia' Turknlenistan'
rezolvarea a patl-tl p'Jf"t"" (problemele 1 4 de
mai jos)' Acestea au fost
problemele 2, 3 dificultate medie. Problema 4 - grad ridicat de dificultate'
cu grad ridicat de
Pentru stuclenli. insa. Problema 3 s-a dovedit a fi cea
Au fost acorciate 9 lredalii cle aur, 18 medalii de argint, 29 de medalii de
..Alexandru Ioan cuza"
maxim: ou,id"iu l,lecrLlai -lt,dianet de Ia universitatea
din Iagi.
prezentim. in continuarc. problemele de concurs gi solutiile acestora,
asemenea. prezentS,m qi soluliile
aqa cum a* fbst indicate de autorii 1or. De
aate ae c5tre unii studengi' diferite de soiu{iile autorilor'

http: //rnath.etti.tuiasi ro/seemous201B/
gamin@fni' unibuc' ro
Universitatea din Bucuregti' Bucuregti' RomAnia'
3)l]niversitatea politehnica Bucureqti, Bucuregti. Romirria, ioana.luca@mathem'pub'ro
4)Universitatea din Bucuregti. Bucuregti, Romania. cornel .baeti.ca@fni.unibuc'ro
5)universitatea Babeg-Bolyai. cluj-Napoca. Rornania. ttrif @math'ubbcluj 'ro
30 ARrrcor,B

Problema 1. Fie ,f [0,1] + (0,1) o func{ie integrabilS Riemann. Ardtali

z *f'l*1a* [' y'g1a,
lo' Jo
l,' oar
* 1)dr t tco'
Vasileios Papadopoulos, Democritus University of Thrace, Grecia
Jurtul a cons'iderat cd, aceastd, problemd, este s,impld. Concurenlii au
confir'rnat tn bund, mdsurd, aceastd, op,inie, mul{t, d,intre et, r'eugtnd sd, gd,seasca
calea spre rezoluare. Juri,ul a cons'iderat cd, partea rnaz delicatd, a solu{te,t este
demon,strarea faptului, cd, i,negali,tatea cer"utd. este strtctd, moti,u pent,ru care
prr,nctajul marim pentru aceastd, problemd, a fost acordat doar acelor studenli
co,re au argumentat tn mod sati,sfd,cd,tor aceastd, chest,iune. Solu[ii,le pe care
dori,m sd le propunem di,ferd tn esenld. doar tn acest punct. Din acest moti,u,
tn loc sd, prezentd,m mai, multe soluli,i, care coi,nc,id la ntuelul detali,i,lor si,mple,
am optat pentru uarianta une'i si,ngure solulr,z, tn cadrul cd,re,ia, clrcstiunea
i,negali,td,lii, stri,cte ua fi tranqatd, tntr-o lemd, pentru care r)orn prezenta patru
demonstral'i'i, pentru a cd,ror ordonare am lrnut cortt de c6,t de elaborate sunt
elementele teoret'i ce uti,lizat e.
Solutie. Faptul c5 orice functie integrabilS Riemann pe un interval compact
ale c5rei valori sunt pozitive are integrala pozitivd este o consecin!5 imediati
a definiliei integralei Riemann qi se va folosi in mod repetat in cele ce urmeazd
fdrd vreo mentiune explicitd.
Incepem prin a proba o lemd care se va dovedi util5 atAt pentru a legitima
scrierea fractiilor din enunt, cAt gi pentru a ardta cd inegalitatea finald este
Lemma L. Fie a,b e IR, a ( b, gi, h: [a,b] -+ R o func{ie i,ntegrabi,ld
R'iemann care are toate aalorile stri.ct poziti,ue. Atunci I ttlrlar > O.
J,, \

Demonstra{ia 1. Presupunem lr(r)dr: 0. RemarcXm c5 in

"d J"fo
situatie h(r)dr : 0 pentru orice c, d e [o,,b] pentru care c ( d, cdci
altminteri am ajunge la contradictia

n1r1o, : h(r)dr * h(r)dr + n{dar ) 0.

l"o f,' l.o loo
in continuare avem nevoie de
Lemma 2. Fi,ea,0 e IR., o ( 0, n e N* fz g:lc-,P) -+R ofunc{i,e
i,ntegrabi,ld, R'iemann care are toate ualorile pozi,tzue qi, i,ntegrala nuld. Atunct

la. pl astfet tncat

Ic g(r) a I nentr"u
eri,std, interuare tnchi,se ned"egenerate
ori,ce r € I.
inchis neriegenerat inclus in [a, 6] va conline elemente
r pentru care 9(r) > ] '
Consider[m qirul de diviziuni (An)qex. ale iui [ri'
6] cu

1-\"'- q
o+zi a^o" B)
(, (€ql. (qz' ' ' ' , {rr), unde pentru fiecare
qi sisternele de puncte intermediar"
€ + (k- l)+'o +k?] aie((sk) >
s €N-ri A'e { 1.2..-..q} avem€nr Lo
corespunz[toare lui 9, A, qi (n va fl
|.q^pentru fiecare Q € N*, suma Riemann
2 e:f s6ol). care este mai mare decAt #,
a" u"de ff g@)d'r' * 'o'
k:1 n
< € ["'b']'
nedegenerat.Il: lor,bt] c [a,b] astfelinc6't h(t) lpentru.oricer
la r6,ndu-i i, ipotezele lemei
Constat6m ca restriclia lui h'la 11 se incadreaz5
nedegenerate (1/l)'ex.
2. continuand incluctiv. construim un qir de intervale
pentru orice r €
astfel incAt In'r : [a,+r.b,+r] C lo,,,bn) qi h(r') < #
Pt .rp{, k i k €N.} e,!n-''' deci h(c) S'1 pentru
lanat,b,+r]. Atunci .
orice n € N*' de unde h(c) :0' conttadiclie'
: existd rn €
!].-n-'(-* +-))'
Demonstratia 2. intrucAt [0,t]

pentru care F-
g, n-1 ([#,+*)) ,,, neglijabild Lebesgue. ExistS, prin
".;" nu este con{inut5, in nicio
urmare un numar a >'d"1., proprietatea ca F-
motiv. oricum a- Iru o diviziune A
: (a : J0' jtl: "''rn: b) a lui [a'b]'
printre sumele Riemann i n((r)(', - ri-r) asociate acesteia
se vor numdra
qi unele care conlin suUe*presil nta,)(ri - ri-r) satisfXcand condiliile 'L c
11,2,...,n), D(r,*t-,,) > Aqi h(€i) > p"'ttt'oricei € L' Acestesume
' deci [! n61a" > *'0' contradiclie' D
Riemann vor fi mai mari rlec6t f
sale de discontinuitate esie neglijabil[ Lebesg,.e.
Cum intervalul [4, b] nu este
(o', b)' Fie c un astfel
negiijabil,h admite puncte de. continuitate pL intervalul
de punct. Atunci exist5 d > 0 astfel incat (c -
6,c*d) : (.,,b) qi h(r)
> dh(c)'+> 0'
pentrufiecer e (c-d,c*d). Seobline [!n61a"> t:::h(r)dz tr

Demonstralia 4. Fiind integrabilX Riemartn (pe interval compact), functia

h este gi integrabilS Lebesgue; cum valorile lui h sunt pozitive qi # h@)dr :
0, deducem cE" h(r):0 aproape peste tot pe [a,b], contradictie. tr
Revenind acurn Ia solu(ia problemei 1, constatdrn, aplicand lema 1, ci
nurnitorii fracliilor din enun! sunt nenuli, cleci aceste fraclii au sens. Avem
r.f (r)2 < f (")'pentru orice r e [0,1); aplicAnd lenia 1 (pun6nd 0 in loc de
l. f Q)2 pentru a ne incadra in condiliile acesteia), oblinem
(o .)l rf @)2d,r . lo'
f {il'a, (1)


1)dr .r lo'
/(r)d.r. (2)

ultima integralS fiind strict pozitiv6 conform lemei 1. Din relaqia (2) obtinem
0< <-l-l
- 2 I; /(r)dr (3)
Jirtal' + 1)dr
iar din (t) qi (e) oblinem inegalitatea doritS. tr
Observalii. (1) in cele precedente s-a probat de fapt inegalitatea din enun!
pentru orice fr-rnclie integrabil5 Riemann "f , [0,1] -+ (0. +co).
(2) La aceastd problemi 5 studenti au obtinut punctaj maxim.

Problema 2. Consider5rn numerele naturale m.t).p.q > | qi matriceie

A e M,.,.,,(R), B € Mr,,e(R), C e IZr,.,,(lR), D e Mq,,,r(IR), ar;a inc6,t
At : BCD, B, : CDA, Ct : DAB. Dt : ABC,
Demonstra\i cd, (ABCD)' : ABCD.
Aceastd problenr,d, nu este orr,ginald, fii.nd datd, irt, anul 2004 la concursul
uniuersitalti, Taras Sheuc.h,enko di,n Llcrai.na. Dzn pd"cate acest l'ucru" a fost
remarcat foarte tarziu, iar juriulu'i i,-a fost adus la cunoEttnld l,rt, dimi,nea{a
concursulu'i, cd,nd era difi,ci,l ca problema sd, fie suasa din listd" qi, tnlocur'td,.
Solulia 1. Cu P: ABCD: AAt e .M-(R.) avem
p3 : (ABC D)(ABC D)(ABC D) : (ABC)(DAB)(C DA)(BC D)
: DtCtBI(BCD): (BCD)I(BCD): (At)t(BCD): AaCD: p.

: p <==+ (p' - P)(P + 1,,,) : O*.
p3 (4)
Vom clemonstra cd det(P+I-) f 0. Presupunem c5, det(P+I,") : 0. Aceasta
implicd existenla lui X e M,,,r(lR) \ {O-,r}, aqa incAt (P + 1",)X : O,m1t.

SEENIOUS 2018 33

: O*,1 €+ llAt + I*)X : O*;

(P + I*)X
+ xt(AAt + I*)x : o +=+ (AtX)t(Atx) : -xtX. (5)

Dar, pentru orice Y : (At,A2,... ,Arr)t eM-,r(R) avem YtY: i A? >- O,

in timp ce (5). in care X * O,n,r,arat5, cd
(Atx)t(Atx): -xtx <0.
Aceastd contradiclie ne asigurS, cX det(P+1-) * O, qi astfel relalia (4) implicd
Solutia 2 (Dragoq Manea, student, Facultatea de Matemati,cd,, Uni,uer-
sitatea Bucuregti). Matricea P: ABCD: AAt e M-(lR) este simetri-
cd ((AAt)t : (At)tAt: AAt), gi astfel diagonalizabil5 (cu valorile proprii
)r,... . ). reale). Deci, existd o matrice inversabilS ^9
e M-(R), aqa incAt

P : Sdiag (,\1,..., )-) S-1. (6)

Pe de alta parte avem

p:pt <+ ABCD:(ABCD)t,

(ABC D1t : DtCtBtAt : (ABC)(DAB)(C DA)(BC D) : (ABCD)3,

deci P3 : P. Astfel, din (6) obtrinem

P3 : P <==+ diag(,\f,...,):") : diag()r,...,)-)
<+ )f, : )r, Y k:1-,...,n'1,
ceea ce implicd )e e {-1,0,1}, V k:7,...,ffi. Dac[ aritlm cb )r # -1,
vom avea A? : )r, gi cu aceasta justificarea, egalitS.lii P2 : P este incheiatS,

,Sdiag ()?,. . .,^?*) 5'-1- Sdiag ()r.... , )-)S-1 <+ P2 - P.

S5. demonstr5m, deci, cE" ),61-1; de fapt, vom ardta cX ,\p ) 0.
Valorile proprii )r, . . . , )- ale matricei P sunt rS,ddcinile polinornului
caracteristic asociat lui P,

d,et(Xln : X* - o1f,m-1 + o2X^-2 - ...+ (-l)'"o,n.

- P)

Aici, pentm fiecare k:1,... )m) coeficientul ok este egal cu suma minorilor
principali de ordin k ai lui P. Consider[m un astfel de minor, care provine
din liniile .jt,...,.7k. RemarcS,m c5, deoarece P: AAt, acest minor se scrie
34, ARrrcor-e


/r,, \
: a"t | , I (il, ,1-) .

\t,^ /
unde -Lr, , . . . .L jo sunt linii ale matricei A. iar ;i reprezinta pro-
dusul sialar euclidian, respectiv norma euclidiana in :-'-'. Folosind faptul
c5 det(I/,l/r) > 0 pentru orice 1J.f e M,,.n(-(.). din l7i rezulta ca minorii
principali ai lui P sunt nenegativi, qi c1 aceasta o;, ) !. A': 1.m' Acum,
presupunind c5 polinomul caracteristic are o radacina ,\ < 0. ajungem la

0:detl^t,r- P) : )* _ ot\- | +o2^"' I - " - r-1\'no^ :

: (-1)," {(-l)"' * or(-.\ r"r-L - ... 1o,n)} + O.
Cu aceasta demonstralia egalit[1ii P2 : P este incheiatS..

observalii. (1) Justificarea faptului c5 det(1,,, - P I este nenul. echivalent,

matricea P: AAt nu are ca valoare proprie pe ) : -1. poate fi fScutd
remarcAnd c[ P este matrice pozitiv semideflnita.

(AAtu.u) : \At'u. Atu\1 : llAtt'll2 2 0. t = .!4,,.1(lR.) ,

Ei apelAnd la rezultatul cunoscut conform cdruia o a-.tfel de matrice are toate

r-alorile proprii ) 0. Aceasti observalie a fost folosita de catre unii studenti
la solu(ionarea Problemei 2.
(2) intr-un spa(iu vectorial V cu prodlts scalar matricea ar.And componen-
tele (r';. i_,r). unde utt...,r1r, €V, se numegte matricea Gram corespunzfltoare
sistemultii de vectori ul :....ur.,. iar determinantul acesteia - determinantul
Gram. Astfel. dacd A € M-,r(R.), matricea --llt este matricea Gram co-
respunzdtoare liniilor lui ,4. iar determinantul din formula (7) este un de-
terminant Gratn. O matrice Gram este pozitiv semidefinita (gi orice matri-
ce pozitiv semidefinit5 este o matrice Gram). in particular. un determinant
Gram (ca produs al valorilor proprii ale rnatricei corespunzS,toare) este nene-
gativ. Aceasta proprietate este apelati in solulia 2 a problernei, prin remarca
det MMt > 0.
Inegalitatea det 11,41t ) 0 se mai poate demonstra folosind formula Binet-
cauchy pentru calculul determinantului produsului a douS matrice.
(3) La aceastd problem[ 22 de stridenli au oblinut punctaj maxim.

SEEN{OUS 2018 35

Problema 3. Fie A,B e y\lzors(iR) cu proprietatea ci AB: BA\i 4.201'8 -

32018 - 1. unde 1 este matricea unitate. Ar[ttrli cd dac6 tr(A) : 2913,
atunci tr(A) : tr(B).
Vasileios Papadopoulos, Democritus University of Thrace, Grecia

Solu{ia 1 (a autorului). Deoarece l2oLB - 32018 - I ai AB: BA averrr

cd" (AB)2018 : L
De aici oblineur cX valorile proprii ale matricelor A, B qi
,48 sunt radacini de ordinul 2018 ale unitdtii. Din faptul ci tr(AB) : 2018
deducem cA valorile proprii ale lui /8 sunt toate egale.:l , (l\'Iai general,

dac[:1.....:2018 sunt numere complexe de modul 1qi I zr,:201.8, atunci

zt:... : :lurr : 1. Aceasta rezulti imediat: scriem ,ll qi
201f 2018 ".o"aplisinal
din f :r. :2018 deducem c5 D cosal.:2018, deci cos ak: I pentru orice
A:1 &=1
k:1.....2018 qi, in consecint5, sinotr :0 pentru orice k - 1,...,2018.)
Aqadar polinonrul caracteristic al matricei AB este P.q6 : (X - 1)zots.
Pe de alta parte. polinomui minimal al matricei AB, pta, divide pe X2018-1,
respectiv pe if - 1)z0ts. deci p.ne - X - l iar aceasta insearnni cd" AB: I.
Astfel am oblinut cd. B : A 1.
Se qtie ittsA ca valorile proprii ale inversei unei matrice sunt inversele va-
lorilor proprii ale acelei matrice. in cazul nostru inversele valoriior proprii
aie matricei -1 sunt egale cu conjugatele lor (deoarece au modulul 1), iar de
aici dedr-icenr ca tr(B) : t(,4) : tr(A), ultirna egalita,te avAnd loc pentru c[
matricea -{ este real5, deci gi urma sa este tot un num5.r real.
Solulia 2. Fie )1.. A': 1,...,2018, valorile proprii ale rnatricei A. respectiv
H*. k : 1. . . . . 2018. r'alorile proprii ale matricei B. Deoarcce AB : BA, :nta-
tricele -1;i B sunt simultan triarrgularizabile (peste C), adicX existd o matrice
inversabilA L- e -\12613(C) cu proprietatea cd, matricele LtALi-| Ai UBU-l
sunt snperior triunghir"rlare. (Acest lucru se demonstreazl prin induc(ie dup5
dinren,.iunea uratricelor observAnd ca lB: BA implicA faptul cd cele dou[
matrice au un vector propriu conlurl.) Aqadar ralorile proprii ale matricei
AB sunt (eventual dupa o renumerotare) l6pr6. A : 1.....2018.
(Se putea. de asemeneal argumenta cd matricele -{ qi B sunt diagonali-
zabile. deoarece valorile lor proprii sunt printre rad5cinile de ordinul 2018 ale
unit5tii, deci distincte, gi atunci sunt sinrultan diagonalizabile.)
Din tr(,48) :2018 se obline. ca la solulia 1. ca \tltt :1 pentru orice
k : 7,. . . , 2018, adicd valorile proprii ale lui B sunt inversele valorilor proprii
ale lui A rsi rre g5,sim astfel in situalia de Ia solutia 1.
Observatii. 1) Solu{iile studcnliior care an rezolvat complet aceast6 pro-
blemd, au fost in spiritul celor doua solulii prezentate rnai sus, diferentele
ap5rAnd doar la nivel de detalii.
2) La aceastd problemd 15 studenli au ob{inut punctaj maxinr.
36 ARucol-R

Problema a. (u) Fie / : ]R. -+ R. o funclie polinomialS,. S5 se demonstreze

(,r)clr : /(o) + /'(o) + /"(o) + ' "
lo* "-'f
(b) Fie / : IR. -+ R o func{ie care aclmite dezvoltare in serie N{aclaurin cu Iaza

de convergen!5 1? : oo. s5 se demonstreze c5, clac5, seria ; /t"l {o) este abso-
lut convergentd,. atunci i,tegraia improprie [- ,-'f
(r) dr este converge.ti

gi are l<,rc egalitatea

e *f(r\a..:I-f""iu).
Ovidiu Furdui, Universitzrtea Tehnic[' Cluj-Napoca' RomAnia

Solulii qi coment arii. To[i concurenli,'i ca're au rezoluat Ttrr'nctul (a) a'u dat
una d,intre urmd,toar.ele d,oui solu{,i,i,. Surpri,nziitor, aIL Jost uni,i' conatren[t'
chr,ar rli,ntre stutlenli,i rorn6"ni,) care au primit 0 puncte la acest ,,ereTci'l'iu d'e

(a) Solulia 1. Fie d gradui lui /' IntegrAnd prin p6rti, oblinem

-' Ilr')l; .
[- "-'l(r)dr : Jo[- ,-" ')'./(r)dr : - ( t -'l'1r;d;'
Jn l-
: f(o)+ | e-'f'(r)dr.
/ sunt tot
RepetAnd integrarea prin pdrli gi {inand seama c6 derivatele 1ui
furrclii polirromiale' se obline

.lu* "-"
f @) a' : /(o) + /'(o) + lo* "-' ,"
('L') dr : '' '

: /(0) + /'(0) + "'+ ('t)

(0) + (r)dr
/ "-r /d'+r)
/(o)+ /'(o) + ..+,f('r)1s;.
deoarece .1(d+t) : 0.

Solulia 2. Deoarece / este funclie polinomial5, conform forrnulei lui

Taylor avem

tlrl:.' J_
) lrnr(0)
/-/ --------- I"
oriCare ar fi r € R,
SEEI\{OUS 2018 37

unde d este gradul lui /. Drept urmare, avem

: :
lo* "-' f
(') d''
fo* "-* "a" *+9 r(n + 1)

Punctul (b) s-a douedi,t a fi cea mai, di,fici,ld, problemd' d'in concurs. El
a fost rezoluat complet doar de patru studen{i. (Oui.di,u Neculai, Aud,danei',
George Kotsouol'is, Georgi,os Kampan'is 9i, Gydrgy Tdtds). Soluliile celor pa-
tru, studenli s-au bazat, tn esen{d,, pe teoremele clasice de conuergenld' (teo-
rema conuergenlei, un'iforme, teorema conuergenlei monotone sau teorema
conuergen[ei dominate). Prezentdm mai, jos solu[i,a autorulu'i, precurn Ei doud
di,ntre solulii,le date tn concurs de studen[i. di,n Greci,a.
(b) Solutia 1 (a autorului). Pentru orice numdr natural n not5m
/(1(o) ,r,
s, :: l,k:0ftkrtol precum gi r,(r),: ik:0
cel de-al n-lea polinom Maclaurin al lui /. Cum Zlo)(O) :1@)(0) oricare ar
fi k e {0,1.... . n}, in baza lui (a) avem
I e-'Tr(r) dr : Sn pentru orice n € N.

Pentru a dovedi cX integrala improprie If f (") dz este convergentd,

vom ardta c5 limita # e-'f
(r) dz existd qi este finit5. Folosim teorema

lui Bolzano. Fie e ) 0 arbitrar. Deoarece."riu i l/(")(0)l este convergent5,

existd un ri.o € N astfel ca
l/(,)(o) | .;
Cum integrala improprie ff dr este convergenti (conform punc-
tului (a)), limita J516 exist5 qi este finit5,. in baza pdrlii de
necesitate a teoremei lui Bolzano, existS un d > 0 in a4a fel incat pentru orice
't),'u' € (6, oo) sX avem

I[ ."' '-.',,(')d'l < ;

l"u I
38 ARrrcor,r.;

Atunci pentru orice u, u' e (6,co) cu u < u', avem


qg l,'' ,-.,,d, -
r+qf(nr1) :
rt: t Lol L

: 1+
2' t./'

Cum e a fost arbitrar, in baza pir{ii de suficientra a teoremei lui Bolzano

rezultd c[ limita,[1# e 'f (r)dr existS Ei este finitd. adic5 integrala im-
proprieIf "-" f(") d;r este convergent5.
lr- e-'.f(r)dr-S,l| : lr* e-''(/(.rt-Tn(r))
lllJ,' I lll/o
drl <


rn ' L-
A.:n I 1


Convergenla seriei impiicX faptul ca girul (S") este convergent,

t l/(k)(0)l
av6nd limita [f, e-'f (r)dr. D

(b) Solu{ia 2 (datd, tn concurs de un student di,n Greci,a). Fie qirul de

funclii 9, : [0, oc) -+ IR (, > 1), definite prin
n t r(k)
grlr)::.-'t +' (o)l _r

SEEMOUS 2018 39

Evident, qirul (g"(r)) este cresc5tor pentru orice z € [0, oo) fixat. IVIai mult,
notand S:: I lyrkttO)1. avem
O<g,(r) (,5e-'Ih t pentruoricen> 1Eioricere [0,oo).
fr:0 =
Prin urmare. existi
)yyS"@) € IR oricare ar fi u € [0, oo). Fie funclia
9: [0,co) -+ ]R. definitS,prin 9(r):: j!1O"(";.
Deoarecegiruldefunc{ii (g")
converge uniform pe compacte, rezultd cd g este local integrabilS Riemann.
In baza teoremei convergentei monotone, avem
IJ g,*)d, : Jo lim g,(r)dr : lim I S,(r) dr.
I n-+x n-oc Jo
r\ n

I s,@)ar: Ik=o l/(fr)(o)l oricare ar fi n ) 1.

conform celor demonstrate la (a). Deducem de aici cE
ut{or; :5.
IJo glrldr: n-+x
-linr f tf

deci /o- 9(r) dr converge.

Fie acurn qirul de functii .f" : [0, co) -+ R (, > 1), definite prin

::"-" / A

f ,@) ik:0 lior ,r

Deoarece / este dezvoltabil5 in serie Ntlaclaurin pe IR., rezult5 ci

,,19 /"{"; : e-" f (r) oricare ar fi z € [0, m)'

Avem gi
f"(*) Sg"(r) ag(r) pentru orice n > 1 gi orice z € [0,oo).
Aplicdnd teorema convergenlei dominate. deducem cX
*f fx f@
l,* u
(r) dr : / lirn (r) dr
n-)- /,
: n+ooJo
lim I f"(r) dr.
f,(r)d,r: t/(fr)(o) oricare ar fi n ) 1,
conform celor demonstrate 1a (a). Drept urmare, avem
,oo n oo

I e-" f(r)ar : Jg5!/tk){o)

: t/(k)(o).
Jo k:o k:o
40 ARrrcor,B

(b) Solutia 3 (datd, tn concurs de un student di,n Greci,a). Se aplic5
fn: l},oo) -+ IR (ri, > 1), definite prin
girului de func{ii

fn,),:.-" /(;lo) ,,.

urm5toarea variantd a teoremei lui Tonelli (a se r-edea R. Gelca gi T. An-
dreescu, Putnam and Beyond. Springer, 2007. p. 178): dacX
roo @ )c f).
,/ f V*@)ldz ( ro sau t
u - 4 /
t 'fn\.rt d.r ( >c.
J\) n:0 n:gJU
atunci are Ioc egalitatea

/-; r*@)dr:tr-|,- l"@)dr.

REmAne doar s5, notim c[. in ipotezele problemei. ar-em
x roo co

I1Jo/ \'' ' ' -''' :Il/'"'o'

rr " < x'
dr : / r f,/r.r r dr'.
f ./, (.r)
J l)
precum qi
t / f,(r)dr:f /"
n o'u n:0
Observa$ie. Pentru Problema 4 punctajul marirn (10 puncte) a fost obgimrt
de 9 dintre studenti.
A. B6ruvr, I. Cegu, A snppv cASE oF MATHEMATICaI (lr,lts)tNouc'rtou 47

A happy case of mathematical (mis)inductionl)
AnpLo B6Nvl2), IoAN Cagulr)

Abstract. From a rnissed attempt to mathenratical induction and through

a journer- in calculus, we arrive to a stronger inequalitv conccrning partial
sunts of the harmonic series.
Keywords: hrequalities. mathematical induction, sequences. series
MSC: Primarv 26406, 26D06, 26D15; Secorrdary 40A05.

The topic of an exercise iri [1] is the following inequality: for all n e N,
n> 2. \\'e have

t*;"('- hl

This folloii's straightforwardly from the AN,I-GM inequality if we re-write it


uA. L.'
, V;
However. a-. n-ith many of the statements that contain the phrase "for all
n € N...". iots of our students will be tempted into using the Principle of
Mathematic'al Licluction to prove it. It does not quite work for this statement
for reasons that n'ill become clear soon, but our story has a happy ending
triggered bv a series of fbrtunate calculus events.
If ri'e clenote by P(n) the statement in (1), one easily checks first that
P(2) is trLre being equivalerrt to ,/, < 1.5). Showing that P(k) + P(k + l)
is at the c'ore of the method atrd, in this case, it comes down to proving that
for all i: : I i:
k+7 k
,ffi-m'iu k

The statement expressed in (2) is quite strong in the serise that the difference
bet*'een the left and right-hand sides is getting srnaller as k gets larger' This
is alreaclr' apparent for small values of k: when k : 1, (Z) is Jl- 1 = 0.'11 <
0.5 : 1/2. ri-hile rvhen k : 2, (2) is ffi - 1/Z x, 0.665 < 0.666 < 213. A
computer algebra s1'stem sucli as Mathematica easily verifies that (2) holds
1)This work is partially supported by a grant from the Simons Foundation (No. 246024
to Arprid B6nyi).
2)Department of Mathematics, 516 High St, Western Washington University, Belling-
harn, WA 98225, USA, arpad.benyi0wwu.edu
3)Departrnent of Mathematics, West I-lniversity of Tirnigoara, Bd. Va^sile Pdrvan, no. 4,
300223 Timigoara, Romania, ioan. casu@e-uvt . ro
42 Norr: N{arRruATrcE

for other randomly chosen natural values of k, so...(2) must be true?! Yet, it
is not an obvious inequality, and the considerations above do rrot amount to a
rigorous proof of it. We extrapolate that a typical student would conclude at
this point that showing (1) bV induction is either not possible or, if possible.
not an easily accomplishable task.
The moral of our tale is that doing matheniatics is a fun roller coaster
and perseverance pays off in the end. We u'ill indicate below how basic
ideas from calculus combine to prove (2). In s-hat foilou,s, we will assume
that k e N and k > 20. This is iust a technicaiitl' that r,vi1l simplify some
of the calculations; as mentioned already, the fact that (2) holds for k e
{1,2,.. . , 19} carr be checked directly by a computing device.
Let f : [1,oo), f(r): The inequalin'u'e q'ant to prove can be
re-written as

f lk +1) - /(tu) . #. v A' > 2(.r (3)

First of all, this inequality is ureaningful in tlie sense that the expres-siott otr
the left is always strictly positive due to the fact that .f ::::::ctlv increasing.
Incleed, since In(/(r)) : (1- *)t"r, by irnplicit diffe:er.:.aii,-,Ili'e find tltat.
for all r ) 7, f'(r) : ,-t-tlr(r - 1+ Inr) > 0. \lore,,,r'er. br- the \Iean
Value Theorern we see that proving (3) is equivaient to prc\-ilrg that for some
0 e (0, 1) we have f'& + 0) < #.Now. a tedious calculation further shows
that, for all r ) 1,
f" (r1 :r-s-tlr((htr)2 - 2lnr * r - 1
- *-3-r1rS.((lnr)2 - Inr + 1i - ,r'- In,,')] > 0.
Thus, / is concave up (that is, // is increa.sittg r. atid so it suffices to show
that for all k € N and k > 20,

f'(k+1) < A1'

A' -+-

This last inequality can be rewritten as

(r+ t)-1-1 l(k+1)(k*ln(k+ t)) < o ln(k+ 1) < k((k* t)1/(t+rl - t;.
K+ -I
ttl.: ffi- I > 0 <+,k: (1- n1.)l <+ P: -ln(1EGI-
+ o1)

With this notation, (4) is equivalent to

I.(1-+ al+r)
( o
(k + 1)ln(l * a*+r)
a[.+1 . -!-
,kar.-r (5)
k+ 1

Intuitively, we see that (5) is just a quantitative version of the '*'ell-known fact
that, assurning we have ak:0, !+#P :
1. In the remainder of
/l[ AL

A. B6wvr, I. Cagu, A uappv cASE oF MATuEMATrcal (urs)rNoucuoN

this note we show that (5) holds for k e N with k ) 20. Our argument will
be based on two sirrrple observations.
The first observation is that. in<leed, o^ : 0, and, more impor-
tantly, we can express this convergence in a ^llrl
quantita,tive way. Wb can show
this by using the Binomial Theorcm, which, in particular, plives that for all
(l * r)k > l + l)
tr., 1k(k r:-.

If we plug in.r: 2lvq iu the last displayed inequality, we see that

/" 2\r >1+-r-k>A'. A-(A"-1) 4_

which gives
o<W-t< +*n^.4'l' (o)

br- the Squeeze Principle

lbr sequences) we obtain '/from herc : 0.

Our second observation is that. for all :l ) S, -,ve have ^.lj1nu

;2 13
ln(l*.rtr-.r -1+
. (7)
The student that has encountered seriers in his or her studies will recognize
that the right hand-side of (7) is just a partial sunr of tlic series expansion
of ln(1 + r). Regardless of this, wo car] prove (7) easily as follows. Let
h : [0, oc) -+ R, h(0) : 0, and. for r > 0.

h(r) :lrr(1 *,r') - z. + - +

:.f- 1+r- : -' ; S0.
r.r' -
h'(r) x2
I .r" r 1 -
h is decreasing ancl hencc l,(r) < 0 for r > 0.
\&'e are readv to irrrplenrent our two observations in obtaining (5). Ap-
plying (7) to r : .ra-1 ri'e get
ln(1 + a611)
<1- ak+t
2- J
, o'k+,

Thus, it suffices to show that

. ak+t ^2
23 k+t
that is
tl (rt. r\
(k+l),,r"+r(.:-;) r,
44 Norp iVIa.'rpuetrcl:

Notice now that, bV (6) and for k > 20. we have

1 47.-,-1
- 1 2 1

2-i>r- zr/E+f i'

We are left to show that for n) 27 we have nan > 3 (, : k * 1 above).
Rewriting, we have

But since (l+yltlu < e for all 9 > 0, letting ?J:31n, we get

\ n,/
for n) 21. The proof of (5) is complete.
An immediate corollary of (2) is that, as observed at the beginning of
this note, we have
-. (\/ k+7 k k \
k-{/F+ 1 - A._ 1 )
ktrm m-
or, equivalently,
,. / k+l k
ii1(rffi-Tk) :r

We note first that for ali k e N, (k+t)t/t't+1) < l'1 ' ,. ''j;! a
in turn follows easily from the fact that the function 9 : x),5@): be it
decreasing; sinceg/(r) :ry (0forr)3. \ori.br i21 andsince(VE; it
a decreasing sequence, we have

1 ;. :
111. -
-*-y[+1 L, k,-:i a : +.
W'{T - iF-V?
and the conclusion follows from the fact obserr-ed before tfrat {/E : 1
and the Squeeze Principle for sequences.

[1] T. Zvonaru and N. Stanciu, Problema 2'/269. Gazeta \,{atematicd (Seria B), no. 9
L.-I. C,q.r,q.N,q,

The monotony in hazard rate sense of some families of

mult idimensional distributions
LutcI-IoNuT C.q.reNat)

Abstract. Irr this note we will prove that the fronotony in hazard rate
sense holds for some families of multidimensional distributions.
I{eywords: Stoc}rastic orclers, risk theory, hazard rate, distribrrtion
MSC: 60E15'

For a random vector X : Q -+ IR.d we consider its distribution p(B)
P(X e B). its distribution function F(r):P(x < r)' and F.('):P(X2t'
x I t). In this article, the distribution function for another random vector
Y u'ill be denoted bY G.
In this note tve are using the notation ancl some well known results from
l1l and l2l.
Definition l. Let X and,Y be two rand,om uectors. we say that x 'is smaller
thanY tn hazarcl rate sense (and' we denote X <h,Y ) if
F-(r)G.(y) < F-(, Av)G.(rv v), Vr,y € IRd'
Definition 2. Let X and, Y be two random uectors. we say that x ::
srnaller than Y in uteak hazard, rate sense (and we denote X {*h. Y ) fr
is i,ncreas'ing on SuPP(G.).
Theorem 3, Let X,Y be two rap,d,om uectors. If X <nrY then X l*h. Y'
Theorem 4. Let X : (X)i_-,rr1 and'U : (Yi)*17 be two ro'ndom uectors'
If X <n Y then Xt <n,Yi. i': l,d''
Theorem 5. Let (Xt)*Ti and (Y1)r-* be random uariables'
If Xt <,n, Y, i' : 13, the, 8!:rX, <r,. ef:rYl'
For the sake of completeness, let us recall the definition for some mul-
t iclinreusional distributions.

The multtuaTiate unzform distribution Unif (1) on some product of in-

tervals I : It x ... x 16 C IRd has the clensity function 1@)
: ffi. Its
marginals are Unif (/i).

1)Departa-errt of Mathematics, Faculty of Mathematics and Informatics, University of

Bucharest, Romania, luig j- -catana@ya-hoo' com
46 NorB Ntllrplrerrcn

The multiuariate Poi,sson distribution of ) : ()a, )a_r,... , )r, )o),

Poisson(,\), where each ,\1 is nonnegative, has the density function

_$, d min( 11 r4'

/(r) :, o?'"*
II ^;^
k:1*, i=0 lit'','
i:1 fi;)'
\ r!
The marginals are Poisson(); + )o).
The multi,uariate normal distribution lf (p. D) has the density function

f (,) : (a.t (2, t))-L ".lt'-.ri'L-'"-P),

where p : (EXt, EXz,. . . , EXi and ! deuotes the covariance matrix.
The marginals are l/(p,;, D).

The bi,uari,ate Bernoulle distribution B(1.p) has densitv

where a; e {(0,0) , (1,0) , (0,1) , (1,1)} and pt are four nonnegative real num-
bers with sum one.

1. MarN RBsur-rs
Proposition 6. Let a, b e IR{ . Then Unif (10. a]) <n. L nif ([0. b)) if and onlg
Proof. \Ve have Unif(I) <5. Unif(J) + Unif([0.o,]) <6. Unif([0,bt]),i :
Conversely, assume a /-b. Then Unif([0.or]) <5. Unif(l0,bi]).1 :7,d..
It follows thar Ef:1Unif([0,ar]) <r,. Af:, Unifl[O,b;]), in other words. we
have Unif(fO,rl) <r,. Unif([O,b]). !
Proposition 7. Let a.P € re1*t. f/Poisson(o) {6. Poisson(p). then ai+
ao 1 ,& * go, i : 7.1. Th,e conuerse h,olds if the marginals are inclependent.
Proof. Suppose that Poissor(a) 3r.. Poisson(B). Then for i -- ln one has
Poisson(altao) ln. Poisson(p;1B6), anci therefore oi*n6 < Ai+ 39. i :7,d".
For the converse. if fc,r each i : 7.d it holds .r, + o0 a Ji - B6 then
Poisson(a; f ao) ln, Poisson(3; +,30). Assuming moreover that both random
vectors have independent marginals, it follows that
Ii:l Poi.tot (ar + )o) tu. I Poisson(/ii 1 )o),

that is, Poisson(a) <5. Poisson(B).


Proposition 8. Let b" p, t' € IRd and D > 0. If l{(fr,I)

<1,. l{(p',1) then
haue zndependent m,arg'inals.
l-t s l_t,. The conuerse holds i,f both dtstrzbutions
Proof. From I'(p. D) <n, l/(lr', D) it follows N(pn,I) <n' I{(p'1!), which

z : tJ, so lhat p'11"tt.

in burn implies that 7-r; l pt,for ail
Conversell,. I l-t'means Ltr I lt'i for each i :7,c1- Therefore, one
has l/(p;, D) -n. l{jr't,f), and hence Ai!:rl'{(pu,I) <n, \f:rN(tt'r,I), i,
it ii ii ii
other words .\'(4. D) <n, l/(lr',I). tr

Propositio n 9. We haue B(7,p) <n. B(1, s) e ff ) max (, X,#,ffi)

Proof. Let X and Y be the random vector for B(1.p) and B(l,q), respec-
tivel1,. If X <6. Y then X l*,h. Y, so that fr i. ir."t"asing on Supp(G-),
o, #(r) < F (g), for all r,y e Supp(G.) with r { A. \Ve specialize the
vectors J. g to conr,'enient values.
For r : (--, --) :
,U (L, i) *" get that ;1 > 1
For r : (-i,i) :
,y (+,j) and (1, ,:
(+,i) *" get that
y > *;+.t:2.3. But it is easv to verify the follou'ing implication:
P4 -
Qt, Qi* % .i :2.1 + 9! > 9_.i : 2.3.
P1, - P; * P,t' Pt Pi
For r: (0.0) ,u : (L, i) *" obtain # > m. ancl this implies
q4 \ q2+q3
pn' pr+pt'
I ) *u* (1. q'
. :qt .-9'o?'' ) .
P4 - \ P2 P3 P2-P;t/
Conversely, it is easy to r,'erifv that F*(.")G.(A) ( F*(r Ay)G*(rv y),
Vz, g € 1R.d.

[1] NI. Shaked. J. G. Shanthikumar. Stochastic Orders, Springer, New York' 2007'
[2] Gh. ZbAganu. lfetode llatematice in Teorza Riscu,hn q'i Actu,ariat, Ed. Univ.
48 PRoeLprrs


Arrthors shoirlcl submit prol>osed problenrs to gmaproblems(Qrms.unibuc.ro.

Files should be in PDF or DVI fortiat.
Once a problem i-s ircceptcd and considered
for publication, thc author will be asked to submit the TeX filc also. Tlie referee
process will usually take betweerr several weeks ancl tt'o rnonths. Soiutions ntay also
be submitted to the same e-mail address. For this is-sue. solirtions should arrive
before 15th of May 2019.

472. Let a,b, c e [0, $] such that a*b* c : ri. Prove the foljowing inequality:
/a*b\ /b-c\ /c-a\l
sina*sinb*sinc22--1lsin( 'L )sittI o ,)sinl-tt2
I \ " / \ - \ )l
Proposed by Leonard Giugiuc, National College Traj.an, Drobeta
Turnu Severin, Romania and Jiahao He, South Chrna Un:.versity of Tech-
nology, People's Republic of Chi.na.

473. Let €tt...,enbe the elcurentary symnletric poll-n',nrials in the variables

ex(Xt,...,Xrr) :
and let ^9 be the ideal generated by €1,....e, iu l.-f- .-\r,.
Then every monomial Xf''..Xn'" with the degree n7: n71l...*mn
strictiy greater than ([) belongs to S. On the other hand. there exists a
monomial of degree ([) which does not belong to S.
Proposed by George Stoica, New Brunswick, Ca:rada.

". r +... \r} -

lrt+*r----.-- 1l
474. Calcutate f{2n - 1) I (
n=r L\n' (n-1\' / " I

Proposed by Ovi.diu Furdui and Alina Slntdm5.ri-a::, Tech:ri-caI Uni-
versity of Cluj-Napoca, Cluj-Napoca, Roroaaia.
475. We say that afunction /: R -+ IR. has the propertv (P) if it is continous
zf (f (")) :3f (r) - r for a1l ;r e R'
a) Prove that if / has property (P) then ll,l[ - {;r e R. : f (r): r} is a
nonempty interval.
b) Find all functions with property (P).
Proposed by Dan Moldovan and Bogdan Moldovan, Cluj-Napoca, Roma-
PH< lt,ost.tD I)ttoBLnt\ts 49

476. Calutlate the integral

/m arctgr ,
Jn {7 + Lo"c '
Proposed by Vasile Mircea Popa, Lucian Blaga University, Sibiu,

477. For every complex matrix,4 we denote by 1* its adjoint, i.e., the
transposed of its conjugate, A* : AT . If A is square and A: A* we say that
A is self-adjoint (or Hermitian). In this case for every complex vector r we
have r*Ar € R. If A, B are self-adjoint we say that A > B if r*Ar) n*Br
for every complex vector r.
For a complex matrix A we denore lAl2 : AA*. Note that 1,412 is
self-adjoint and ) 0. (If A*r : A : (h,.. . ,An)r then r"lAl2r : U*A :
(at,... ,an)(at,... ,un)r lal2 + ''' + la"l2 >- o.)
Let A be a square matrix with complex coefficients and 1 the identity
matrix of the same order. Then the following statements are equivalent:
(i) 11+ zAl2 : lI - zAl2 for all z e C;
(ii) 11 *zAl2 )Ifor allze C;
(iii) ,4: o.
Are these statements still equivalent if we replace "complex" by "real"
Proposed by George Stoica, New Brunswick, Canada.
478. Determine the largest positive constant k such that for every a' b, c ) 0
with a2 + b2 + c2 :3 we have
(a+b + c)2 + kl@ - b)(b - c)(c- r)l S g.
Proposed by Leonard Giugiuc, National College Traian, Drobeta
Turnu Severin, Ronania.

479. Let p be an odd prime number and A e Mp(Q) a matrix such that
det(An * Ir):0 and det(A + I) * 0. Prove that:
a) t(,a) is an eigenvalue of A + Ir.
b) det(.A t Ii :
-det(,A - Io) fu t) a\(A) + 2.-
Proposed by Vlad Mihaly, Technical University of Cluj-Napoca,
Cluj-Napoca, Ronania.
480. Let k,nbe natural numbers, rL,fr2,... ,fik be distinct complex numbers
and the matrix Ae M*(C) such that (A -r11")(A-*zln)"'(A-r6ln):
O,.,. Prove that rank(A -rJn) *rank(A -rzln) +"'+rank(,4, -npl,-):
n(k - 1).
Proposed by Dan Moldovan and Vaslle Pop, Technical University of
Cluj-Napoca, Cluj-Napoca, Romania.
50 Pn.oei,sNis

481. Let 1r be a field and let n ) 1. Let A,B e LI,,(I{) such that IA,B)
commutes u,-ith A or B.
If char lf : 0 or char K > n then it is knou,n that lA. B] is nilpotent,
i.e.. iA,Bl" :0.
Prove that this result no longer hoids if 0 < char I{ I n.
(Here b1' 1.,'] u,e mean the comrnutator, [A.Bl : AB - BA.)
Constantin-Nicolae Be1i, IMAR, Bucharest, Romania.


455. Let n ) 2 be an integer. Determine the largest number of real solutions

the equatio, ar/i16+...+ a,r1/7lbu:0 can have. HeIe &tt... jan ate
real nurnbers, not all zero. and b1 , ...,bn are mutnallr-distinct numbers.
Proposed by Marius Cavachi-, Ovldius University, Constanla,

Ed'itor's note. Unfortunately, the solution plor-ided by the author is

r rrong. A reformulation of the problem is the follor-ing:
Given ri ) 1 'uve rn'ant to find the largest m for which there are mutually
distirrct bt,...,br, mutually distinct ix7t....r. arrd not all zero &71 ...,an
such tlrat Lnrl/.,'t+bi -:0 for 1< z < m. Equivalentll,, if A e ,4.f-.r(1R.),
A: (\E+4)..3, therr lhe equation AX:0 has a non-trivial real soiution,
viz., X: (ar, ...,ar)T. l'or this to happen one neecls that rank A I n - l.
Hence rn is the largest rrurrrber such that there are mntualll- distinct bl, . . . ,bn
and mutualh. distinct 2:1,...)rr77u sltch that the rank of the rn, x n matrix
A: (\/Gi+, hj)i; is < n - 1. Obviousl1,.. when nr ( rr - 1. regardless of the
values of r; and t.,;, x.e harre rank A { m.( n - 1. so it is po-ssible to have rn
solutions. The author claims that r,r,hen 'm,: n. i.e.. I is a square matrix, we
always have det A+0 so rank A: n, so it is not possible to irave n solutions.
Therefore the rerquired maxirnum is n - 1.
N{ore precisell.. if we rearrar}ge the ri's and br's such that 1r ( . .. 1r,

on tlre fact that the function./ : l0.oo) -+ [0,-), f (.r): v6.ir strictiy
concav€. Unfortunatell,, rvhile the proof is verl, ingenious. it contains an
error. It seems that the result is true urhen n { 3. \\'e don't know what

456. Let .f ,g,h be non-negative continuous frinctions or1 i0,1] satisfying the
inequalitv f (tr + (1- t)g) > gt(l)h,l-r(y) for all r,;q € [0, 1] and some (fixed)
PRoposBo PRoBLEMS 51

pl v7
1 € (0.1). If. in a<lditorr. we havc I 11lt')d.r:JtI h@)dr: 1. prove rhar,
rr .lo
;1r;a" > r.
Proposed by George Sto!.ca, University of New Brunswick, Saint
John, New Brunswick, Canada.

Solution by the author. \\'e assume first that g,h are positive. not
merely non-negzrtir-e on [0,1]. Then y,+ [i g(s)ds and y ,+ I& h(s)ds are
strictly increasing ancl differentiable lirnctions sending [0.1] to [0,1]. We
denote bv u. r' : 10.1] -+ [0,1] their inverses. Then u,tr are strictly increasing
arrd clifferentiable rvith u(0) : u(0) :0 and z(1) : u(1) :1.
\\e har-e:

: h(s)ds: r.
l,^o e(s)ds lo"o'
Bv differentiating rve get

'u,'(t:)g(u(r)) : t,'(:r)h(t,(r)) : t. (1)

Define u.'(r) : tu(r) * (1 - t)u(r,). Along rvith u and u, ur : [0.1] -+ [0,1]
is strictly increasing and differentiable rvith u(0) :0, u,(1) : 1. Using the
inequality bettveen the arithmetic and geometric means, we have
w' (r) : tu'(r) +
t)ri (,r) > (u' (r))t (u'(t ))'-t .
(1 - (2)
Fionr (1) and (2), the hvpotheses and tlie change of variables ,s : ?r,(",-), it
follows that
71 1l
Jo ,ttt)cls> / /(r'(r)) 'u,'1.r;dr

a [' n' @(r))h'-'(,'(r)) (u'(r))t(u'(r))1-tdz


: [' (u'(r\g(u(rrr) ' (u'\.r)g(u(x)))t

\u\ ta.r: l. 'dr:
If g.h are only non-negatir-e then for any 0 < d < 1we define ga(r):
(1 - d).q(r) * d arrd h5(r) : (t - ri)h(r) + d. Same as for g,h, we have
17 pl
I O;t.r|rt., : t h5(.r)dr:
1. Howev€r, 9d,h5 are positive, not merely non-
Jo Jo
negative. Then. bv coutirruitr-and the compactriess of 10,1] x [0,1], from
l(tr + (1 - i)y) > gt|)h'-'(y) we get that for any r ) 0 there is some
0 < d ( 1small enough such that f(tr+(t-t)y) *e > gi@)hl-'(y)
V0 r:, y < 7. Then, by applying the case we have just proved to 95, h5 ancl
f, f i:r6, we s"t J;(/(") +e)da; > 1, i.e., .fi t@)a, *e 1. since rhis )
happens for cvery e ) 0, we get our result. tr
LtZ PRoeLsNrs

457. Lt:t A, B e M,,(C) bc so that A2 + 82 + A* B:2(AB + 1,,). Prove

thr. follt-ru-irrg eqrralit ies:
a) Tr((A - B)(A- B + 1,,)) : 2r.
b) det((A - B)(t- B+ 1,,)):2".
Proposed by Vasile Pop, Tehni-cal University of Cluj-Napoca,
Cluj -Napoca, Roman:.a.

Solutzon bu tltr: author. We use the notatio" i' ] for the commutator
[x, ]'] :XY -YX.
The matrix A,'I: (A- B)(A- B + 1,,): (-{ - q2 + A- B u,,rites as
LI : A2 + 82 - AB - BA+ A - B : (AB - B-\) +21,, : lA,B)t21,,.
Since Tr[A.B] :0 we get TrIVI: Tr(21,,):2n. i.e.. we have a).
Let C : A B, so that lI : C (C */,). \ote that [C, B) : lA- B, B) -
lA, Bl. Then '*,e have
C 1r'+ 1,,) : lC. B)
* )t (3)

and equivalently
lC.B] -C'2+C'-21.. (4)
If we note D : IC,B) : C2 + () +1,, then D commutes u,ith C. \\'e prove
that 'I\'(Dfr) :0. k: Ln.Inrleed. s'e have
Dk _ Dk_t.D: Dk tlCn _ nC)
- Dk tcB - - c(Dk tBi - tDk-rB'1c,
so that Dk : iC,Du 'B), which implies fr (Dft) :0. In conclusion

Tr(D) : f. (D') : -' : Tr(D"t : 0.

so that, if ,\1.)2,...,),, are the eigenvalues of the nratrir D. frorn the svstem
tl + x] + ... + \?*: o.

we obtain )r : lz : " r : tr, : 0 (frorn the Newton forrriulas the poly'nornial
/p with the roots )r, )2,..., ),. is unique, more precisell-/p(A) : )".) From
tlre equality (3) it results that the eigenvalues of tlie ntatrix C (C +.f,) :
D +21, are p1 : lt2: "': F,n:2 and their prodpct is
det (C (C + In11 :2",
i.e., the eclualitv b).


Note. Tlie method to prove that )r - ' r ' : )r, : 0, i.e', that
ttsetd liere
l) is nilpoterit, calr l.tc usecl to prove a rrrore general rcsult: If X,Y are square
rnatriccs of thc same climertsion n and [X, Y] commutes witir X then [X, Y] is
niJ.potetrt. Inotrr casc X : C.Y: B antl we have that [C, B): C2+C-2In
comrmrt,es with C. The resrtlt holr1s, with the same proof, over every field K
of charactelistic 0. If the c:haracteristic is p ) 0 then it holds only if n < p.
See problem .181 florn the current issue of GNIA.

458. (Corrected) Fol a continuous and non-negative function / on [0,1] we

deflne the Hausdorff moments

LL, i: /t r"y1..1.1.,', z : 0, 1,2,. . . .

o ,t

Provc that
t-tr+zt lt| -f p,,,,p,1 ) 2ptylarl.Ltltot fr : 0,1,2,. .. .

Proposed by Cezar Lupu, University of Pittsburgh, Pittsburgh,


Solut'ton by th,e auth,or. Let us recall thc celebra,ted Schur's inequality

rn(,r-y)(:r- z)+a"(y- r)(y -r) + z"(z-")(r- g) > 0 (5)
for all fr,l1, z ) 0 and n, ) 1. Then we have
.r ,: I I1t I1l (X -r Y t Zlf lL tJ'ty)l(:)drdydz ) 0.
Ju Ju Jo
wlrereX ::f," (" - y)(.r - z'),Y -- y"('!l - z)(y - z) arrd Z : z"(z-r)(z-'a).
we lrave ,I : ,Ix -t Jv -r J2, wttet'e.tu : #,C'# xf @)f (y)f (z)d,rd,yd,z
and similarly for ./v. .I 7. But, by s"-vrrrmetrY reasons, Jx : Jv: Jz. Thus
3Jx : J > 0, so ..fu ) 0. We have
11 17 rl
Jx : Ittl I I 7r"+2 + :r"yz - r"*ry - *"+r r)f (r)f (y)f(z)drdydz
Jo Jo .lo
1t ft fl
I fttidy JoI fGl:
I r'-2f111d..,.Jo
- Jo
ft 1t rt
I .r" !1.r1c).r I ttf fy)rls / zl'1:)d:
Jo Jo .lo
11 tt fl
- I ,,'"'rf\.rt,lr I g,l'\tt)tlg I ftrla,
Jt .lLt Jrt
;1 11 1l
- lo ., ''-1l1.r I f ls),ta
ldr J,," " ,Jo/ :/(z)d:
: l.Ln +z ll} I Lo * 1t," 1t.y 1-11 - lln+t - LLn+t l1,O llt t
l-Ll l-to

so 1t",r12p,f,1 p,,.lt'i - 2lr,,,a1ltsl.1,t : Jx ) 0. Hence the conclusion. I

54 PRosr,oN4s

Solution by Leonard Gi.ugi,uc, Nati,onal College Tra'ian, Drobeta Tut-nu

Seuerin, Roman'ia. If f :0 then we have nothing to prove. So we assume
that / I 0, which implies that all lli are positive.
The relation to prove writes as
Pn+z.lto- n .!!-,2,

Pn+l l-tt Pn+t lto

By the Arithmetic X.fean Geometric Mean inequality rne have
F:)2 . ro t ttn . ltr Z, @.
pnil ltt pn*7 2 l_to \l l1i+t
Hence it suffices to prove that p,n12pr, > F?,+t. But this follows from the
Cauchy-Bunyakovsky-Schwarz inequality:
/ ft \ z rl \
Fn*2t1n - (/ r"'2f ir)dt) (/. f t(r)dt')

(1,' "n+zf(r)rnf(r)dr )': (/;' ,n-, rt )dr) : F2n,,.

The proof is complete. n

459. The faces of arr icosahedron ale colored ri ith blue or white such that a
blue face cannot be adjacent to more than trr-o other blue faces. What is the
Iargest rnmber of blue faces that can be obtained folloiving this rule?
(Two faces are considered acl.jacent if ther- share an edge.)
Proposed by Eugen J. Ionaqcu, Coli:mbus State University,
Columbus, GA, USA.

Solution by the au,thor. \\ie prove that the ans\\-er is 1.1 and the coloring
that gives this maxirrrum is the one shornn in Figure 1(a). t'hich is a Schlegel
diagram of the icosahedron in u,hich the projection is done from a point close
to a blue face. (That is. there is an extra blue face tirat cannot be seen.)

Figure 1(a) Figure 1(b)

Figure 1. Icosahedral graph
PRoposBo PRoBLEMS 55

Indeed, for the each of the 20 faces fi, Tz,. .. ,T2s we assign a Boolean
variable tri,i:1,...,20, which is equal to 1if the face I is blue or 0 if the
f.ace Ti is white, given a certain coloring. Let us denote by I/, the indices j
for which f.aceTi is adjacent to 4. The condition we require is then written

rt,* D "r( 3, i : 1,...,20. (6)


Clearly, we want to maximiz S : f ,0. After summing all these

" I
inequalities up over i : 1,.. .,20, we obtain S + 35 ( 60. Hence, it follows
that S ( 15. Since we have arrangement that accomplishes 14 blue faces we
only need to prove that 15 blue faces are not possible to be arranged without
violating one of our requirements.
So, by way of contradiction, let us assume that it is possible to have
15 blue faces satisfying the requirement of a 2-dependence set (as in the text
of our problem). Then all the inequalities in (6) have to be equalities. So,
for every face that is white we need to have exactly three blue around it and
for every face that is blue we need to have exactly two around it. Then we
are forced to have a coloring as in Figure 1(b) if we start with a white face
(the one in the middle) and then we end up with too many white faces (at
least 7), which is in contradiction to the number of faces left possible (i.e.,
20 - 15:5). tr

460. Let X be a set with at least two elements, and fix a,b e X, a I b. We
define the function f : X3 -+ X by

flr,a.r) : ( alf r,y,z*a,
t b ff a i {r',a,r}.
Is there a binary operation ',: X2 -+ X such that f (r,A,z): (**y)* z for
all r,y,z € X?
Proposed by George Stoica, Uuiversity of New Brunswick, Saint
John, New Brunswick, Canada.

Solutzon by the author. The answer is no. Indeed, let us assume that,
forer-ery- r,ye X,r,af a,we havez*A*a. Then,forevery r,y,z€X,
r,U, z f a. '*,e have
a : f(r,A,z) : (* * A) * z f a,

a contradiction. Hence, there exists u,u e X, u,tL f o, such that z *'t): a.

FYom here it follows that

a)Fa: (u*u)+a: f (u,u,a):g


and then, for every r e X we have

b * r : (a * a) * r : f(a,a,r) : 6.

a : .f(b,u.u) : (b* u) * u : b* u : b,
a contradiction with a + b.

461. Let A,B e M"(R) so that A2: A, 82: B, and det (2A+B) :0.
Prove that det (A + ZA1 :0.
Proposed by Vasile Pop, Tehnical Universi-ty of CIuj-Napoca,
Cluj-Napoca, Romaaia.

solutton by Francisco Perd,orruo and Angel Plctza. Departamento d,e lr,ta-

temdti,cas, Uniuersidad de Las Palmas de Gran Canaria, Espafi,a.
If det(2A+B) : 0 then there exists u I 0 in lR." such that (2A*B)u : 0.
From (1 - A)(2A+ B)u:0 and (I - A)A:0 ne get (1 A)Bu:0 and -
so Bu : ABu. Similarly. from (1 - B)B : 0 and (I B)(2A* B)u : 0
we get Au : BAu. Then (A -l2B)2u : A2r: -f2ABu -f2BAu * 4B2u :
Au *2Bu * 2Au * 4Bu: (3A*6f')u and so (A+28)(A+28 31)u:0,
where 1 is the identity matrix.
It (A+28 -31)u : 0, then (A+2f)u: 3t,. Together with (2A+B)u :
0, this implies that Bu :2u. But B is idenpotent so its only possible
eigenvalues are 0 and 1. Hence (A + 28 - 31)t, I 0. Thus det(,4 t 28) : g,
becarrse (A+28}u : 0 has the non trivial solutio, (A+28 - 31)u in 1R.,.. x

The atfihor's solution is essentially the same up to the relation (-4 f

2B'Sz, : 3(/ -l2B)u : 0. Here he assumes that clet(.A + 28) I 0 and
concludes that (A *28)u:3u. Together rvith (2-,1 -r B\t:0, this implies
that Au : -'u. Hence A2u: -Au:,u. But l2L': Ar,- -u. so u:0.

Soluti,on by Moubi,nool O,marjee, Lyc€e Henri IV. Paris, France.

We prove that if A+28 € GI"(R) then ker(2,4 -F B) : g.
Take z e ker(2A + B). We have 2Ar * Br : 0. i.e.. Br :
When we multiply to the ieft by A we get ABr : -2A2r : -2Ar : Br.
When we multiply it to the left by B u'e get that B2r: -2BAr. But
B2r: Br: -2AX, so Br: -2BAr arrd Ar : BAr.
Then (A+2 B)2r : (A2+ZAB+2BA+48\r : Art2Br*2Ar*4Br :
3(/ + 2B)r. Since ,4 -t 28 e GI,, (R.) we have (A + 2B)r : 3r. We mrrltipll-
to tlre left by,4 and we get (A+ 2AB)r, : 3Ar, rnhence Ar : ABr : Br.
Since also 2Ar -t Br :0, we have Ar : Br: 0 and so 3e : 2Ar * Br : 0.
So we proved that if A + 28 € Gr"(lR) then ker(2,4 + B) : {0}, i"e.,
By contraposition, if det(2A + B) : 0 then det(4 * 2F) : g. I

PRoposen PRoBLEN4S 57

462. If .f , [0,1] -+ IR is a convex function with /(0) :0 then prove that

'r(1,' ,r(r)d'r - lo''' /(')o') lu''' rr(r)dr'

Proposed by Florin stdnescu, $erban cioculescu school, GSegti,
D5mbovila, Roroania.
scthtt,ton by the o,uthor. we need two well known properties of convex
Lernma f. if 1isanintervaland /:1-+IR.is aconvex functionthen
for every a e l the function ro :.I\{r} -+ IR.. r,,(r) : tkFIL, is itrcreasing'
Lemma 2. If I is an interval arrd / : 1 -+ IR. is a convex firnction then
for cvery r,?J,z,f € l with r <A 1z <f rzu'e itavc
fty)-fl,) _ /(1)-,/(:)
A-t - t-;
B1' Lemma 1 vr,itir a : 0 the frrnction P ,r increasing. It follows thrit
for every' z e (0. tl2l, a e l},ll2)we have + < $f. ..,
rf (r) +'yf (r) < rf (r * v) Vt,.11 e 10,112).
(When r: 0 we have equalitY, 0 : 0.)
we integrate this inequality first u,ith respect to g. then rvith respect
to r. We get

),.r,*,+ jtt,r ,- l, ./(.r+ y)da:*

f ,'-" fr,,n, tr. [,;]

'' '/,'iot) n'

; 1,"'rl(r)cr-r -l lr" l(,,',u,
= lo''(;',) (1,'

:! r )dr -; 1,"' ,' (, (, - ;) - rtr)) 'r r

Here in the right hand side of the inequtrlity we used iritegration bv parts
ancl the fact that the derivative of r+ ir+tlz f (t)dt is /(r +112) - f (r).
\\'e now use Lernma 2. If r'y e [0,1/2] with r < y then 0 I :r I y I
r (o *;) -t (".;)
(,. ;) - (,. ;)
which implies that /(r +112)- f (r) < f @+112)- f (y).Hence the function
r ,+ f (r + 112) - f (r), r e [0,1,12), is increasing' Then, bv applying the

Chebyshev inequality U! t(t)g@a t > fi .t t(tat Il g(ilat if / and s have

the same monotony) we get

(' (". 1) - r1,;)

n,> ,
fr't',ra, lr''' (t (,.;) - y1,,;) o,
f,,,''' "
i(1,: ,/(r)dr - lo''' rt,).,)
In conclusion, we get

'= [''' rf(r)d.r

2lo t"'
-:8Jo s1r1o, l' ./(r)d.,
/ rl n1/n \
2a\Jiz" fe)d.r-l
.lo /(r)dr)
and equivalently
r1/2 pt rtl2
12 I r/(.r)dr1zl
.lo /(r)dr'-2/
Jt1z"' Jo
When we divide by 12 we get the required result. tr

Solutzon by the edttor. There is an alternative approach using a general

property of convex functions. If a < b we denote br-C([a,b]) the set of all
continuous functions on fa, b] with values in R. Then C([4, b]) is a vector
space over IR and it has a metric topology, u'ith the distance given by the
norm ll . ll, where ll/ll : max,,€la,bl
Lemma. If T : C([a, b]) -+ IR. is a linear continuous operator then
T(f) > 0 for every convex function f ,lo, b] -+ R if and on1;, if the following
(1) 7(1) : r@) : 0. Equivalently, T(f) :0 for any' affine futtction /,
i.e., when ./ has the form /(r) : mtr + c.
(2) T(f ,) ) 0 Vc e (a, b), where
,, \ [0. r' e c).
J'"(r):tr_" re fr.bl.

This result is already known but we couldn't provide a reference. For

the sake of self-containment we sketch a proof here.
For the necessity note that if / is affine so is -/. Hence f , -f are
both convex and we have T(f) > 0 and -T(f) : T(-f) ) 0, i.e., f (f) : O.
Since 1 and z form a basis for all affine functions, in order that 7(/) : 0 for
every affitte function on [a,b] it sufliccs to have 7(1) : T(r):0. Since /"
are convex we also have 7'(/") ) 0 Vc e (a,b).

PRoposBo PRoBLEMS 59

Before proving the reverse implication, note that fbr every c e (a,b),
/" is continuous and it is differcntiable evcrywhere erxcept at r: and wc liave
f'(r):0if r< candft(r):1if iL>c. Alsonotethatl"(4,) :Q.
Assume first that / is a convex function a,rrci its graph is a lrroketi line
with the vertices at (c6,do),...,(c,',dr), whele a: (:0 < q < "'4 c:.r: I
is a division of the interval [o,b]. Let ttr; : 'FH bc the slope of the
graph on the interval lct-t,cil. Since / is cclnvex wc ltavr: rlz,1 ( " ' { rn".
The function / is continuous ori fa, b] and diffcrerttiable evcrywlterc except
dt cs,....c-.. Namely, if 1< i < s then .f'(r): Ini.Yr € (c; 1,c1). We ciairn
J@):-r(,-a)+ds1 (tn2- mt)J",(") + " * (nz" rn,-1)f,,",,(r).
If we denote by g(r) the right side of the above equalitv therr the easiest way
to verify that / - g is to prove that g has the same properties with f : g is
continuous, g(a) : /(a) and g'(r) : f'(r) Yr e [a.b] , t f c11,....c.. T]re
contirruity is obvious sirrce each /" is contintrous. Since J,,(a):0 Vc e (a,b)
we have g(a):m1(a- a)+d,s- d,o:,/(c0) :./(a). For the third corrditiort
let r € la,b), r I co,...,c". Then ci-1 < r < c,; for sorrle 1 ( I ( s, so
f'(r) : rni. we have
f'(*) : mt l (mz -,nr)f|.,(r) + " * (rrr, - nr,,-),f '"" ,(:r).
Ifj < i - 1 then c, I c;-r ( -r'. so J!,1.?') : i. Ifj > I tlren t'i )> t'; > r'. so
f '.,(") : o. It follows that
g'(r): mtI (rnz- rrr) + "'+ (rr, - rr"i-,):rfti: f'(")
By the linearity of 7 we get
arj):T(ms(r -o) + do)*(m2-rrt1)T(f1,) + t (rn, -rru, 1)T(f"",,).
But rne(r-o,)+d6 is affine. so by property (1) T(mo(g-a)+do) :0, and by
property (2) T(f.r) ) 0 for 1 < i < s - 1. BuL u''e a,Is<-r have m,1 t "' { Trlst
so each m,i - mi*L is non-negative. We conclude that 7("f ) 2 0.
Suppose now that J , lo,b] -+ R is convex arbitrary. For each n ) 7
we define the function f,, : [n, b] -+ lR. whose glaph is the broken lirre rvith
vertices al (o,+;(b-a).f la+;(b-a)))u'itlr U < i < n. Notc that o -;lb- o)
n'ith 0 < i, < n make a partition of [4, b] into n er1ua,l intervals. Also note
that (a + *(A - o,),f \a+ *(b - u))) are points on tlie gra,ph of f , which is
convex. so /,, will bc convex. as well. Since the graph of f, is a broken liue,
by the particular case we have .just proverl, we havc T(f ,,) > 0.
Since / is corrtinuor:s on a compact interval, it is rrrrifbrrnll, contintrotts.
Then for an). t > 0 there is n., suclt thtrt for alty n ) rr. we have

1f@) - f@)1 < e Yr.y €la,b) rvith lr' - yl 1!1t,- o;

60 PRoeLsl.rs

Letr e la,bl. Then r e lo * *fU - a),a * ;\b - a)lfor some 7 { i, { n.

of [a* *tb- a),at *(U- o)) is ](b-a) we have lr - (a+
Since the Iength
*tu-o))1,1"- @+fi(b-a))l < *(b a), so l/(r) f (a,++(b-o))1, lf @)-
f("+*(u- o))l < e. It foilows that f("+*tu-a)),f(a+ fi(b-o)) €
(f (r)-e , f (r)+e). But r belongs to the intervai la++(b-a), at i(b-a)),
where /, is affine. It follorvs that f"(r) takes an intermediate value betr,r,een
I,@+ *O -a)) : f (a+ f ta-a,)) and I,b +;(b-a)1 : [(o + )(b-a)).
since /(a +*(u -o)),f (a+ ft(b- a)) e (f (r) e ,f (r) +e), this implies
f"(r) e (f (")-e,f (r)*e), i.e., 1f"@)-f (r)l < a. Inconclusion llf"-fll-
ma-x. l/"(r) - /(.r)l < E.
lim f" : 1. Since 7
]lf" - fll < e. for n ) n- u'e har.e rr-+oo
is continuous, this implies :
JIr1f11,, T(f\. But 7(/") ) 0, Vn, so
7(/) > 0. tr
We now return to our problem. \Ve must pror-e that T(.f ) > 0 for every
convex function [0,1] -+ R with /(0) : 0. u-here
"f '
r(r)::6 ( t'
[''' /r, ra,)/ - ro| '
- Jo rr(r),b.

If we try to prove that T(f) ) 0 for all convex functions / we see that this
is not true. We have 7(1) : -* I 0, so the condition (1) of the Lemma is
not satisfied. Therefore w-e need the information that /(0) : 0. The idea is
to find a ) e IR such that TU) > 0 for every convex function,f ,[0,1] -+ IR.,
where T(f) : rU) + )/(0). If we prove that 7 has this property then for
a convex function / that has the additional propertr.that /(0) : 0 we have
T(f) : rff) +^/(0) : 7(/), so TU) > 0 implies T(f) >0 and we are done.
Whenwetake f =7weget7(t) :7(1) +)-1: *+f,soinorder
that ?/(1) : 0 ne need to take ) : *, so T(f ) - T(f ) + *f tOl \\,e have:

: ; (1,',1d2 -
l,''' ro,) - lo'''r:crr * i, : o,

: i (|,',,rd.r - :
l,''',o,) - lo''',2d* +i
.o o,

so the condition (1) of the Lemma is satisfied.

We now prove the condition (2), i.e.. that 7(/.) ) 0 Vc e (0.1). u,here
f.(*):0 ifr < c and f.(r): r- cif r) c. First note that /"(0):0, so
: :
T(f + *r"tol T(f .) vc e (0, 1). we have tu,'o cases:
Case 1. c < ll2. Then on the interval lll2,1l we have f,(r) : r - c,
while on the interval l0,7l2l we have f ,(r) :0 for r e [0, c] and /"(r) : r - c
PRoposop PRoBLEI\{s 61

for r € lc,ll2). It follows that

/ 11 rtl2 \
T(f ,,) : r(fi :u1 (/,,r,, r1l2
- c)dr - J. (.r - c)a.r) - J" ,r(r - c)dr
: - cI3 c2

6 - n+ E: -hrnr,
+2" - 31.
But we have 0 I c .-7f 2, so -& < 0 and 4c2t2c-3 { 4(1142 +2Q,12)-3 :
-1 < 0. Thus ?/(/") : --uU"2 + 2, - 3) > 0.
Case 2. c> 712. Thenon 10,1,12) we have f.(*):0, while onl7l2,7)
welra,ve f"(r):0if z€ lll2,c) and/"(r) :r,*cif re [c,1]. Weget
T(f,,):r(f"):;t/rt t/c2 1\ (c-1\2
(/" (r-c)dr-o)-o:; (; -"*;): ? rn.
Hence 7 satisfies ihe conditions (1) and (2) of the Lemma, which irnplies
thatT(.f) ) 0 for every convex / : [0,1] -+ R., ancl we are donc. n

463. Prove that there exists n6 € N such that Vn ) v1o the equation

-l-*.L*...+ 1 :rnrz
l*r 2lr n*r
has a unique solution in the interval (0,-), cienoted by 2,,, and that
,\,!'y'n, - o,
where a e (0, 1) is the unique solution in the interval (0, 1) of the equtrtion
, i" - 1, where 1 is the Euler constant.
Proposed by Dumitru Popa, Ovidius University, Constanla, Romania.

Soluti,on by the author. For every natural number n let us define 7, :

1+ + +...+ L - lnn. As it is well-linown,0 ( 1r, .--7, Vn € N, and
:1€ (0,1). For every nai,ural number nlet hn: [0,oo) J R,
It,,(.r) : #+#* .+*+" lnn. Let us note that all h,, are continuous and
strictly decreasirrgon [0,m). We have h,.(0) : 1+]+ +*-lnn: ?, ) 0,
h,,11; : i+i+...+#-Inn:1,,.-l +#.
Since,,l5gir,,1t; -
1 - i < 0. it follows that there exists n6 € N such that Yn, ) ns we have
h,,(1) < 0, Tlms there exists ns € N sr.rch that Vn, ) n6 the equatiort
h,,,(n) : 0 has a unique solution in the interval (0, *) and this soit-ttion,
denoted by ,n, is in the interval (0,1), that is 0 ( r,n l l,Yn ) no. For
evely n ) n11 wt ltave hn(;:,,) - 0. T+ + r+ +...-t- -]- : lttn. or
62 PRosr-sNrs

(r+i+ +*l -(++7i+ +"h) -i,.andrhus

.I-l )
* 11,i a *S: 1n Yn ns- (/,1

For every natural number n define gn: {0,1] + iR b1'

p,(r)-,f+, -ln :ir1---r

=i\i-;*r)-1'' ) -
Define also p : [0,1] -+ IR. by
p(r) :.L, *+;-1 - t (i\ - *) / -,
i-l i:l
Let us observe that Vr € [0, 1]. Vz e N 0 S,h-+,, < * and the series
D i Ir convergent, by the Weierstrass
i:l i:t
criterion. so the series ! ,z{; it
uniformly convergent on [0, 1] ancl its sum is a continuous function, that is,
p is a continuous function. o , is a strictll- increasing function. Since
equation p(r) : 0 has a unique solution in (0. 1). ri'hich is denoted in the
sequelbya€ (0,1). N{oreover, since foreverl-n € Nandr € [0,1] it holds
"-*Li(i+rl *Li(i
*r) ' / ___(
* \- l(i+r)-'
,L ii2'
i:l i=t i:n.t i_nll
it follows lim q"(r) : p(r) uniformly rvith
that ro-+cn '
respect to z € [0,1]. From
the uniform com€rgence. by a general result. rnhich rl,e rvill prove in the end
of the proof, it follows that 1im [er,(r,)- p(r")):0. Since by (7) Yn) no
gn (rn) : 0, we deduce that g@") -+ 0 : g(a). Since ,p is strictly increasing
it follows that rn-+ o. Indeed, let e > 0. Take 0 < ry < min(e. a.l-a):such a
numberexistssince0 <a < 1. Then0 < a-q <a <a+71 < l and.sincerpis
strictly increasing, q(a-q) < p(a): 0 < p(a+q).From p(r,) -+ 0 it follows
that there exists n, ) ns such that Yn ) ne , p(a-n) < p(r") < p(a+4), that
is, since rp is strictly increasing, Yn ) nu, a-n < rn I alT,|xr- al < n < €.
rn: a.
It remains to show that if ,{X f"(r) : p(r) uniformly rn'ith respect to
r € [0,1], then for every sequence (rr,)r.ex c [0,1] it follows that
j5g[r"("") - p(*")) : o. Indeed, we have: VE > 0, 1n.. € N such that
Yn ) nu and Vr e [0,1] we have lp"@) - p(r)l < e . In particular. for
PRoposnp PRoBLIJI\4s

fr : fin we deduce that Vn ) ri. we have lpr(r".) - p(r")l ( 6, i.e.,

J$[r, (r, ) - p(r, )] : o. !
64 INsrRuc'rIUxI pEt{'t'RL AIrTonI

Instruc!iuni pentru autori

I\{aterialele trimise redac(iei spre publicare vor fi r'edact:rte intr-una dintre lin-r-
bile romAnfi, englez5, sau li'aricez5. Articolele riLl vor deplqi 8-10 pagini, iar notele
natematice qi cele metodice 2-3 pagini. Autorii sunt r'.tg:r{i sd, fragmr:ttteze ei iirglqi
materialele de dimensiuni mai mari perrtru a sc irrcadla iir lirtritelc inc'iicate mai sus.
Fiecare articol va fi insolit de un scurt t'ezttrnat aI lucr':rrii. nr:de'pirgind i5-20 de
randuri tehnoredactate in lirnbile francezX sau englezi. de inclicii de clasillcare ANIS
gi de cAteva cuvinte cheie. Notele de subsol vor' fi inclrtse in pagina trlde se facer
Referinlele bibliografice se vor trece 1a sfArqitul 1uci5lii. confbrur nclrtuelor stan-
dard internaliorrale. prescrrrtirile adrriise liind cele uzrtale. fblositc in N{atliernatica,l
Orice mamrscris va fi iriso{it de adresa exactd a arttor'r}ui.
N{anuscrisele vor fi. de regulX. tehrroredactate in lirnbajul LaT'eX. Fiqierul sursS gi
cel pdf vor fi transmise prin e-rnail la adresa of f ice@rms . unibuc. ro sall expeclitrte
Ia adresa redacliei.
Orice material se va publica pe rdsprrndelea autoluhii. r'edaclia neasumAndu-qi
fondul opirriilor exprimate. Nlanuscrisele nepublicate nri se restituie autclrilor.
I Pentru lucrd,rile cu caracter original. copl'r'igirt-ul trplir'line ediiorului. Orice te-
producere integralS sau parliali a rrnui material de]ri p,.rblicat ,qe va lace nttrnai cu
acordll scris al SocietXlii qi cu specificarea coordortatelol exa.cte ale priuiei apari{ii.
Autorii pot comanda 1a reclaclie, contra cost. extrir-i('dirt rer.istti conlinand arti
colele publicate.
Instructions for authors
The articles submitted to the editorial office shoukl br: u-ritteu in Romanian,
English or French. The articles u,.ill not cxceecl E-10 pages and tlie mathertiatlcal
class-room notes 2-3 pages. If the articles are lalgel than the above iirrrits. thc
authors are required to divide the articles by tliernseir-cs in order to cornpll' u,ith the
above. Each articie should be preceded by a short sulnnrirr]' of its contcnts in English
or French, not exceeding 15-20 typed rou,s. Kel.rvorcls trnd ANIS classification should
also be attached.
Footnotes should be included at the bottorn of the salne page.
Bibliographical ref'erences shoulcl be givcn at the enci of the trrticle by usirrg the
international standard norms and the adrnitted abbleviatlorrs accordirrg to lvlathr:-
matical Revier'vs.
Any manuscript should be acc:ornpanicd b1. the exact addrcss of thr,' author.
The manuscript should be processed in LaTeX ancl the sollrcc fiie and the pdf file
should be mailed to the editorial office by usual rnail ol to of f ice@rms.unibuc.ro.
Any material is pubtished in the author's own responsibilit.v. Tire editorial board
will assume no responsibility for the contents of anv prrblished items.
Unpublished manuscripts are: not to be returned to the authors.
For any original paper the copyriglrt belongs to the pLihiisher al} riglrts reserved.
Any rnaterial already published may be reprocluc'ed. in r'r,hole ol irr pa,rt, onl1, u,ith
the written permission of the Society trnd be ruentioning tlie exact dates ol ihe first
The authors can order, b.rr pavment, r'eprints of theil ilrticlcs.
Paralela 45 Pub lishing House recommends

Radu Eologan :,: i ,

l,n S'.lxn Mariiei.n

IIII uinat uoaea noMANla

Ro man ian

€ 6bdF Ailma kteh CARAGEA


,Eg PABrrEg


Trparul executat la tipografia Editurii Paralela 45

SUMAR - CONTENTS - Vol. XXXVI (CXV), nr. 1-2, 2018


1. Computing exponential and trigonometric functions of matrices in

Mz(C), by Ovidiu F\rrdui t
2. Stirling type formulas, by fon-Denys Ciorogaru t4

5. A happy case of mathematical (mis)induction,
by Arpdd B6nyi and Ioan CaSu. ........ 4L
The monotony in hazard rate sense of some farnilies of
multidimensional distributions, by Luigi-Ionu! Catana. .... 45




@ Materialele primite Ia redaclie (probleme, note, articole etc.) se publicH

cu acordul implicit al autorului care, odatil cu transmiterea spre publicare,
cedeaz{ s.s.M.R. dreptul de proprietate intelectual{ asupra acestora.
Copyright-ul apartine editorului (S.S.M.R).
Toate drepturile privind reproducerea, parlialx sau total[, sub orice
forml a materialelor publicate in Gazeta Matematic5, suat rezervate
Societ5tii de $tiinle Matematice din Romfrnia.

ISSN 1584-9325 25 lei