J
OURNAL DE
T
HÉORIE DES
N
OMBRES DE
B
ORDEAUX
J. C. L
AGARIAS
J. O. S
HALLIT
Linear fractional transformations of continued
fractions with bounded partial quotients
Journal de Théorie des Nombres de Bordeaux, tome
9, n
o2 (1997),
p. 267-279
<http://www.numdam.org/item?id=JTNB_1997__9_2_267_0>
© Université Bordeaux 1, 1997, tous droits réservés.
L’accès aux archives de la revue « Journal de Théorie des Nombres de Bordeaux » (http://jtnb.cedram.org/) implique l’accord avec les condi-tions générales d’utilisation (http://www.numdam.org/conditions). Toute uti-lisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte-nir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
Linear Fractional Transformations of Continued Fractions with Bounded Partial
Quotients
par J.C. LAGARIAS ET J.O. SHALLIT
RÉSUMÉ. Soit 03B8 un nombre réel de développement en fraction continue
03B8 =
[03B10, 03B11, 03B12,... ],
et soitune matrice d’entiers tel que det M ~ 0. Si 03B8 est à quotients partiels bornés, alors
[a*0, a*1, a*2,...]
est aussi à quotients partiels bornés. Plus précisément, si aj ~ K pour tout j suflisamment grand, alorsa*j ~
| det(M)|(K
+2)
pour tout j suffisamment grand. Nous donnons aussi uneborne plus faible qui est valable pour tout
a*j
avec j ~ 1. Les demonstrations utilisent la constante d’approximation diophantienne homogèneL~(03B8) =
lim sup q~~ (q~q03B8~)-1.
Nous montrons queABSTRACT. Let 03B8 be a real number with continued fraction expansion
03B8 = [a0, a1, a2,...],
and let
be a matrix with integer entries and nonzero determinant. If 03B8 has bounded partial quotients, then
[a*0,
a*1,
a*2, ...
]
also has bounded partial quotients. More precisely, ifaj ~
K for all sufficiently large j, then| det (M)|(K
+2)
for all sufficiently large j. We also give a weaker bound valid for all with j ~ 1. The proofs use the homogeneous Diophantineapproximation constant
L~(03B8)
=lim sup q~~(q~q03B8~)-1.
We show thatdet(M)|L~(03B8).
1. INTRODUCTION.
Let 0 be a real number whose
expansion
as asimple
continued fractionis
and set
where we
adopt
the convention thatK(0) =
+oo if 0 is rational. We say that 0 has boundedpartial
quotients
ifK(O)
is finite. We also setwith the convention that
Koo
(8) =
+oo if 0 is rational.Certainly
Koo
(8) ~
K(8),
and is finite if andonly
ifK(B)
is finite.A survey of results about real numbers with bounded
partial quotients
isgiven
in[17].
Theproperty
ofhaving
boundedpartial quotients
isequiv-alent to 0
being
abadly approximable
number,
which is a number 0 suchthat
in which
Ilxll I
=min(x -
x)
denotes the distance from x to thenearest
integer
and q runsthrough integers.
This note proves two
quantitative
versions of the theorem that if B has boundedpartial quotients
and Mb
is aninteger
matrix withJ
det(M) #
0,
also has boundedpartial quotients.
co+d
The first result bounds
K ( °’B+6 )
in terms ofKoo(8)
anddepends only
cO+d
on
THEOREM 1.1. Let 0 have a bounded
partial
quotients.
aninteger
matrix with0,
thenThe second result upper bounds
K( de+b
in terms0 )
anddepends
PTHEOREM 1.2. Let 9 have bounded
partial quotients. If
M =[a dJ
isLc
d.
an
integer
matrix withdet(M) #
0,
thenThe last term in
(1.4)
can be bounded in terms of thepartial quotient
ao of
9,
sinceTheorem 1.2
gives
no bound for thepartial quotient
ao
:=I i
ofChowla
[3]
proved
in 1931 that2ad(K(O)
+1)3 ,
a result ratherweaker than Theorem 1.2.
We obtain Theorem 1.1 and Theorem 1.2 from
stronger
bounds that relateDiophantine approximation
constants of 9 and which appearbelow as Theorem 3.2 and Theorem
4.1,
respectively.
Theorem 3.2 is asimple
consequence of a result of Cusick and Mend6s France[5]
concerning
the
Lagrange
constant of 0(defined
in Section2).
The continued fraction of can be
directly computed
from that for8,
as was observed in 1894by
Hurwitz[9],
who gave anexplicit
formula forthe continued fraction of 20 in terms of that of 0. In 1912 Chitelet
[2]
gavean
algorithm
forcomputing
the continued fraction of from that of0,
and in 1947 Hall
[7]
also gave a method. LetM (n, Z)
denote the set of n x ninteger
matrices.Raney
[15]
gave for each M= a
b
EM (2, Z)
with-c
d ’
det(M) =1=
0 anexplicit
finite automaton tocompute
the additive continuedfraction of from the additive continued fraction of 0.
cO+d
In connection with the bound of Theorem
1.1,
Davenport
[6]
observed that for each irrational 0 andprime
p there exists someinteger
0 a psuch that 0’ = 0
+ ~
hasinfinitely
manypartial
quotients
an(B’) >
p. Mend6s France[13]
then showed that there exists some 0’ = 0 + Phaving
p
the
property
that apositive
proportion
of thepartial quotients
of 0’ havean (0’) 2::
P.Some other related results appear in Mend6s France
[11,12].
Basic factson continued fractions appear in
[1,8,10,18].
2. BADLY APPROXIMABLE NUMBERS
0 =
~ao,
al,...]
]
is determinedby
and for n > 1
by
the recursionThe n-th
complete quotient
an of 0 isThe n-th
convergent
v’ of 0 isqn
whose denominator is
given by
the recursion q_ 1 -0, qo - 1,
andan+lqn + It is well known
(see
[8, §10.7])
thatSince an+1 + 1 and qn, this
implies
thatfor n >
0.We consider the
following
Diophantine approximation
constants. For anirrational number 0 define its
type
L(O)
by
and define the
horraogeneous Diophantine approximation
constant orWe use the convention that if B is
rational,
thenL(0)
=L~(B) _
+oo.(N.B.:
some authorsstudy
thereciprocal
of what we have called theLa-grange
constant.)
The best
approximation
properties
of continued fractionconvergents
give
The set of values taken
by
over all 0 is called theLagrange
spec-trum
[4].
It is well known thatL~ (B) > J5
for all 0. If 0 =[ao,
at,a2, ... ~,
then another formula for
Loo(6)
issee
[4,p. 1].
There are
simple
relations between thesequantities
and thepartial
quo-tient bounds
K(O)
and cf.[16,
pp.22-23].
LEMMA 2.1. For any irrational 0 with bounded
partial quotients,
we haveProof.
This is immediate from(2.2)
and(2.3).
0LEMMA 2.2. For any irrational 0 with bounded
partial
quotients
Proof.
This is immediate from(2.2)
and(2.4).
0Although
we do not use it in thesequel,
we note that bothinequalities
Since from some
point
on, this and(2.4)
yield
Next,
from(2.1)
we haveHence
Let K =
K,,. (0).
Then for all nsufficiently large
we haveso
We conclude that
3. LAGRANGE CONSTANTS AND PROOF OF THEOREM 1.1.
An
integer
matrix M= La
d
I
withdet(M) #
0,
acts as a linear[c
oj
fractional transformation on a real number 8
by
LEMMA 3.1.
If
M is anintegers
matrix withdet(M) =
~1,
then theLa-grange constants
of
0 andM(O)
are relatedby
Proof.
This iswell-known,
cf.[14]
and[5,
Lemma1],
and is deducible from(2.5).
DThe main result of Cusick and Mend6s France
[5]
yields:
THEOREM 3.2. For anyintegers
m >1,
letThen
for
any irrational number0,
and
Proof.
Theorem 1 of[5]
states thatLet
GL(2, Z)
denote the group of 2 x 2integer
matrices with determinant~1. We need
only
observe that for any M inGm
there exists some M EGL(2, Z)
such that 1Vl M =0 d’
with a‘d’ = m and 0 b’ d’ . For if0
Ud
so,
and 0 =
then Lemma 3.1gives
whence
(3.4)
implies
(3.2).
To construct MTake C = and Then
gcd(C, D) -
1,
so we maycomplete
this row to a matrixM
EGL(2, Z).
Multiplying
thisby
a suitablematrix
cyields
the desiredM.
1
0±l ’
yThe lower bound
(3.3)
follows from the upper bound(3.2).
We use theadjoint
matrix
-We prove
by
contradiction.Suppose
(3.3)
werefalse,
so that for someM E
Gm
and some 0 we haveThis states that
which contradicts
(3.2)
for0’,
sincedet(M’)
=det(M)
= m. 0Remark. The lower bound
(3.3)
holds withequality
for some values of 0and not for other values. If for
given 0
we choose an M EGm
whichgives
equality
in(3.2),
so that = thenequality
holds in(3.3)
for B’ =
adj(M)(0).
However,
if =V5,
as occurs for 0 = thenfor all
M;
hence(3.3)
does not hold withequality
when m > 2.Proof of
Theorem 1.1. Theorem 3.2gives
:5
Nowapply
Lemma 2.2 twice toget
To obtain the lower
bound,
we use theadjoint
JSince
I det (M) I =
thisyields
4. NUMBERS OF BOUNDED TYPE AND PROOF OF THEOREM 1.2
Recall that the
type
L(8)
of 0 is the smallest real number such thatTHEOREM 4.1. Let 0 have bounded
partial quotients.
an
integer
matrix withdet(M) #
0,
thenProof. Set 1b =
Suppose
first that c = 0 so thatI det(M)1 = ladl
> 0.Then
L(0)
>1,
whereWe have
For any e > 0 we may choose q in
(4.2)
sothat ~ ~
L(1f;) -
e. ThenLetting
c -~ 0yields
(4.1)
when c = 0.We have
so that
We first treat the case qa - pc = 0. Now
since det =
det(M) #
0. Thus if qa - pc = 0 then
lPd -
1,
.J
hence(4.5)
gives
It follows that qa - pc
,-~
0provided
that~/
We next treat the case
when qa -
pc54
0. Now from the definition of we seeGiven e >
0,
we maychoose q
sothat ~ ~
L(~) -
e, and we obtain from(4.6)
and(4.9)
thatimplies
thatMultiplying
thisby §
andapplying
it to the left side of(4.10)
yields
q
Letting
e -7 0 andusing q >
1yields
provided
that(4.8)
holds. Now(4.8)
fails to holdonly
ifThe last two
inequalities imply
(4.1)
when 0. 0Proof of
TheoremApplying
Theorem 4.1 and Lemma 2.1gives
which is the desired bound. 0
Remarks.
(1).
Theproof
method of Theorem 4.1 can also be used todirectly
prove the boundsof Theorem
3.2,
from which Theorem 1.1 can beeasily
deduced. The lowerbound in
(4.14)
follows from the upper bound as in theproof
of Theo-rem 3.2. We sketch aproof
of the upper bound in(4.14)
for the case9 =
with 0. For any c* > 0 and allsufficiently large
q* > q*
(E* ),
we have
We
choose q
= qn(1b)
forsufficiently
large
n, and note thatas n -~ oo,
since 0
is irrational. We can thenreplace
(4.9)
by
(4.15),
andthen deduce
(4.11)
withL(O)
replaced by
+,E*.Letting
q -+ oo,f -+ 0 and E* -~ 0 in that order
yields
the upper bound in(4.14).
(2).
For agiven
matrix M consider the set of attainable ratiosBy
Lemma 3.1 the setV(M)
depends only
on itsSL(2, Z)-double
cosetTheorem 3.2 shows that
It is an
interesting
openproblem
to determine the set BothI
andI lie in
V(M),
as follows from Theorem 3.2 and the remarkfollowing
it.Acknowledgment.
We are indebted to the referee forhelpful
comments andreferences,
and inparticular
forraising
the openproblem
aboutV(M) .
REFERENCES
1. A. Baker, A Concise Introduction to the Theory of Numbers, Cambridge University Press, 1984.
2. A. Chitelet, Contribution à la théorie des fractions continues arithmétiques, Bull. Soc. Math. France 40
(1912), 1-25.
3. S. D. Chowla, Some problems of diophantine approximation
(I),
Math. Zeitschrift 33(1931),
544-563.4. T. W. Cusick and M. Flahive, The Markoff and Lagrange Spectra, American Math-ematical Society, Providence, RI, 1989.
5. T. W. Cusick and M. Mendès France, The Lagrange spectrum of a set, Acta Arith.
34
(1979),
287-293.6. H. Davenport, A remark on continued fractions, Michigan Math. J. 11
(1964),
343-344.7. M. Hall, On the sum and product of continued fractions, Annals of Math. 48
(1947),
966-993.
8. G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers, Fifth
Edition, Oxford University Press.
9. A. Hurwitz, Über die angenäherte Darstellungen der Zahler durch rationale Brüche,
Math. Ann. 44
(1894),
417-436.10. D. E. Knuth, The Art of Computer Programming, Vol. II: Seminumerical
Algo-rithms, Addison-Wesley, 1981.
11. M. Mendès France, Sur les fractions continues limitées, Acta Arith. 23
(1973),
207-215.
12. M. Mendès France, The depth of a rational number, Topics in Number Theory
(Proc.
Colloq. Debrecen,1974)
Colloq. Soc. Janos Bolyai, vol. 13, North-Holland, Amsterdam, 1976, pp. 183-194.13. M. Mendès France, On a theorem of Davenport concerning continued fractions,
Mathematika 23
(1976),
136-141.14. O. Perron, Über die Approximation irrationaler Zahlen durch rationale" Sitz.
Hei-delberg. Akad. Wiss. XII A
(4. Abhandlung) (1921), 3-17.
15. G. N. Raney, On continued fractions and finite automata, Math. Annalen 206
(1973), 265-283.
16. W. Schmidt, Diophantine Approximation, Lecture Notes in Mathematics, vol. 785,
Springer-Verlag, 1980.
17. J. O. Shallit, Continued fractions with bounded partial quotients: a survey, Enseign.
Math. 38
(1992),
151-187.18. H. M. Stark, Introduction to Number Theory, Markham, 1970.
J. C. Lagarias
AT &T Labs - Research, Room C235 180 Park Avenue, P. (J. Box 971 Florham Park, NJ 07932-0971, USA email: jcl@research.att.com
J. 0. Shallit
Department of Computer Science University of Waterloo
Waterloo, Ontario N2L 3G1, Canada email: shallit@uwaterloo.ca